Massive Mimo Handbook 3rd Edition

third edition
Massive MIMO
Handbook
Extended version
ericsson.com/massive-mimo
Massive MIMO Handbook
Extended version
This book contains two documents:
• Massive MIMO Handbook

• Massive MIMO Handbook — Technology Primer
third edition
second edition second edition
Massive
Massive Massive MIMO
MIMO
MIMO Handbook
Handbook
Handbook
Guidingthe
Guiding theindustry
industryon onhow
howtoto
Technology Primer
makethe
make thebest
bestuse
useofofMassive
MassiveMIMO
MIMO
ericsson.com/massive-mimo ericsson.com/massive-mimo
EN/LZT 4/28701-FGB1010987 Uen Rev C

© Ericsson AB 2024
Ericsson | The Massive MIMO handbook Overview 3
third edition
Massive
MIMO
Handbook
Guiding the industry on how to
make the best use of Massive MIMO
2 Ericsson | The Massive MIMO handbook
Introduction
The handbook provides a guide for how
to get most value of Massive MIMO in a
mobile network
Purpose
The main purpose of the handbook is to provide a guide for how to use Massive MIMO to
meet the performance requirements in a 5G mobile network. It should also provide a guide
for how to choose suitable products in typical network deployment scenarios. The handbook
shall also briefly explain key aspects of how Massive MIMO works and how the different
technology components affect network performance in field.
Target readers
This handbook targets primarily the Massive MIMO stakeholders in the service provider
organizations. It can also be used by internal Ericsson organizations.
Scope
The document focuses on Massive MIMO solutions, as a means for meeting the
performance requirements in the network. These solutions comprise time division
duplex (TDD) on mid-band spectrum, ~2.3-5 GHz, frequency division duplex (FDD) on
mid-band spectrum, ~1.4-2.7 GHz and TDD on high-band spectrum, also known as
millimeter wave (mmWave), ~24-60 GHz. Conventional remote radio solutions are also
included as an alternative where Massive MIMO is not needed or is not cost efficient.
Furthermore, the focus is on the radio solution, including radio hardware, antenna, and
Massive MIMO features. To keep the document focused and limited in volume, the baseband
hardware solution, site solution other than radio parts and the antenna (e.g. power,
enclosure, cooling, etc.), transport solutions (backhaul and fronthaul) are not included.
The service in focus is mobile broadband (MBB) as this is the dominating service in all
mobile networks. Developing services, e.g. fixed wireless access (FWA) are also considered.
Services and technologies which are candidates for introduction in the near future are
introduced in a separate chapter.
Differences between the 1st edition and the 2nd edition

In the 2nd edition, information on FDD and mmWave is added. A discussion on the expected
future development in terms of new services and new technology components affecting
Massive MIMO is also included.
Authors team
Project manager: Peter von Butovitsch
Editorial team: David Astely, Peter von Butovitsch, Erik Larsson
Authors: Henrik Asplund, David Astely, Mats Buchmayer, Peter von Butovitsch, Thomas
Chapman, Sebastian Faxér, Mattias Frenne, Christer Friberg, Farshid Ghasemzadeh, Bo
Göransson, Måns Hagström, Billy Hogan, Ylva Jading, George Jöngren, Jonas Karlsson B,
Erik Larsson, Martin Ljungberg, Jing Rao, Jonas Rosenberg, Yu Yang
Review team: Bo Hagerman, Kjell Gustafsson, Raquel Herranz, Ulf Forssen, Håkan Olofsson,
Peter Moberg, Tomas Sandin, Sibel Tombaz , Vimar Björk, Anders Holmgren, Thomas Kinch,
Jan Fredander, Michael Hannemann
Ericsson | The Massive MIMO handbook Introduction 3
Executive summary
— Massive MIMO to meet network
requirements
N
Massive MIMO explores the spatial domain by using

beamforming, spatial multiplexing and nullforming
Requirements
• Coverage
• Capacity
• User throughput
Massive MIMO
This handbook demonstrates how a service provider can use key capability of Massive MIMO is to improve coverage on new
A
Massive MIMO to meet the 5G network requirements of today and higher 5G frequency bands, e.g., enabling similar coverage
and in the future. New use cases and continuous traffic growth on 5G mid-bands as for existing 4G bands using the existing site
require considerably better user experiences (in terms of speeds grid.
and latency) and substantially higher network capacity. Service
providers can evolve their mobile networks to meet these assive MIMO offers higher capacity and a better user
M
requirements with ease-of-deployment and in a cost efficient experience than a conventional solution using remote radio
and energy efficient way with Massive MIMO. units (RRUs) and passive antennas. Massive MIMO can carry
increased traffic growth over a longer period of time.
assive MIMO provides the means for improved coverage,
M
capacity and user throughput of mobile networks by exploiting
the spatial domain. This is achieved by using multi-antenna
technologies such as beamforming, nullforming, and spatial
multiplexing (MIMO), that take advantage of channel and
antenna array properties.
Executive summary
—the network reality
No change
Traffic load
Dense urban high-rise

Inc
rea
sin
g Urban low-rise
tra
ffic
loa
dp
er Rural/suburban
sit
e
sites
Service providers need to evolve their networks to meet the The deployment environment is another important aspect
future performance needs in a cost- and energy efficient way. to consider in the network evolution process. Deployment
In this process, it is vital to understand the network reality and environments are typically classified according to rural/
how it affects the choice of hardware and software solutions. suburban, urban and dense urban. Key attributes of the
Each site is unique in terms of traffic load, expected traffic deployment environment include inter-site distance, type of
growth, deployment environment, ease-of-deployment, etc. antenna deployment (e.g. mast, rooftop or street level), and
spread of users in the vertical domain (e.g. low-rise vs high-rise
Current traffic load and expected traffic growth for a site are buildings). The deployment environment corresponds to one
important parameters in the network evolution process. Many of of the axis in the figure above. Note that sites with different
today’s networks have a traffic load distribution over sites that traffic load characteristics are found in all types of deployment
look according to the figure above, where red, blue and green environments.
correspond to sites with very high, moderate, and low traffic
load, respectively. Hence, most sites experience low traffic load The network reality is central in the network evolution process
and rather few sites have very high traffic load. The traffic load and also key for deciding how to develop the hardware and
per site varies also over time. For example, a site covering a train software product portfolios.
station can be extremely loaded during commute hours, but
lightly loaded otherwise.
Executive summary
—Massive MIMO radios and features
Capacity Compact Coverage
The Massive MIMO portfolio comprises Massive MIMO Compact is for sites where there are constraints on the
radios (hardware) and Massive MIMO features (software), deployment. It prioritizes TCO and ensures that mechanical
and it is designed to meet network performance requirements properties are in line with site constraints, such as size and
(coverage, capacity, and user throughput) as well as constraints weight. These products still provide substantial performance
on ease-of-deployment, cost efficiency, and energy efficiency gains compared to conventional RRUs in many deployment
for all site types. The design is guided by the site classification scenarios.
discussed in the previous slide, as well as specific requests from
the service providers. A high-level mapping of these segments into different site
types can be found in the next slide. It needs to be highlighted
The Ericsson Massive MIMO radio solutions are divided into that capacity, coverage and compact refer here to product
three segments: segments, and should not be mixed up with network properties
such as coverage, capacity and user throughput. That is, there
Capacity covers the most capacity-demanding sites and is not a one-to-one mapping between the segment names and
provides superior performance in all deployments ranging ‘corresponding’ network properties. For example, products from
from dense urban high rises to rural areas. It supports the capacity segment will provide superior overall performance
large bandwidths, high output power and many radio including coverage, capacity and user throughput.This mapping
chains facilitating superior horizontal and vertical domain is used for TDD mid-band, FDD mid-band and mmWave. There
beamforming. is, however, a broader portfolio available for TDD mid-band,
as this is the mainstream 5G deployment option, whereas
Coverage targets deployments with large inter-site distances,
Massive MIMO for FDD mid-band and mmWave have currently
such as suburban or rural areas with a smaller spread of users
somewhat more narrow applicability.
in the vertical domain. This segment has typically fewer radio
chains to lower the total cost of ownership (TCO) compared to Massive MIMO features are developed continuously. They can be
the Capacity products. deployed at any time providing the Massive MIMO radio solution
supports it. The Massive MIMO features are developed to meet
the relevant requirements over time in alignment with the entire
ecosystem. For example, coverage features are necessary in
early deployments, whereas high-capacity features become
more important over time.
Executive Summary—finding suitable

Massive MIMO solutions
Executive Summary—finding suitable Notes p
Change
Traffic load
Massive MIMO for capacity
Capacity Remote radio due to higher

Massive MIMO cost efficiency where traffic
Compact
load is low
Remote Coverage
radios
Massive MIMO for coverage sites
In the network evolution plan, the service provider must For sites with low traffic load and low projected traffic growth,
choose products that meet the requirements with respect to the remote radio may provide sufficient capacity and in some
performance, energy performance, total-cost-of-ownership, cases a more cost efficient solution.
ease-of-deployment and other factors.
The figure provides a simplified view of how the radio product
Generally, sites with high traffic load need Massive MIMO segmentation, Capacity, Compact and Coverage, maps to sites
radios capable of large bandwidths and capacity features like with different characteristics. For sites in dense urban areas with
multi-user MIMO (MU-MIMO), sites with a large spread of users tall buildings, products in the Capacity segment are in general
in the vertical domain (e.g. high-rise buildings) benefit from more suitable, either Massive MIMO where the traffic load is high
vertical domain beamforming, and areas with large inter-site or remote radio where the traffic load is low. In suburban and
distance require radios with high effective isotropic radiated rural areas, where the inter-site distance is larger, the Coverage
power (EIRP) to ensure coverage. At the same time, some sites product segment is the best fit.
have constraints on ease-of-deployment, e.g. size and weight of
radios, that need to be considered.
Outline and reading guidelines
The purpose of the handbook is to describe how Massive MIMO Chapter 3 explains how to design the Massive MIMO solutions,
can be used to meet the network needs. It does not require i.e. Massive MIMO radio and Massive MIMO features, to meet
deep technical understanding. For those interested in Massive the 5G network needs. Firstly, there is a discussion on how to
MIMO technology, please refer to the Technology Primer [TP1-9] segment a radio portfolio that covers all network needs to a
and the book “Advanced Antenna Systems for 5G Network sufficient degree with a limited number of products. Secondly,
Deployments” [1], which describe many key concepts in detail. it is shown how the Ericsson portfolio is segmented to deliver
capacity, coverage, user throughput, cost efficiency and energy
Chapter 1 introduces the concept of Massive MIMO. Short efficiency for different relevant deployment scenarios.
explanations of what it is, how it is defined, why it is useful and
how it works are given. The main abilities of Massive MIMO to Chapter 4 introduces a guide for how a service provider
enhance capacity, coverage and user throughput are established. can choose radio and feature solutions to meet their needs,
The possibilities for Massive MIMO to improve total cost of specifically when (typically) to use Massive MIMO and when
ownership (TCO) in some scenarios are also introduced. a conventional solution could meet the needs in a better way.
Chapter 2 describes the needs of an evolving 5G network. It is Chapter 5 outlines a few services and technologies which are
established that there are general performance needs such as candidates for introduction in the future and that have relevance
capacity, end user experience and coverage (on higher frequency for Massive MIMO. How, where and why Massive MIMO may be
bands). These needs vary across the network, and they also vary deployed are also addressed.
over time. There are also some constraints to consider, e.g. cost
efficiency and energy efficiency. It is shown that there is a good
match between the 5G network needs, specifically, the needs
with respect to capacity, coverage and user throughput, and the
Massive MIMO capabilities and that this is the main reason for
the high interest in Massive MIMO in 5G networks. The network
needs provide input to be considered when designing a product
portfolio that shall meet all network needs. Chapter 2 also
discusses key characteristics and differences of FDD mid-band,
TDD mid-band and mmWave Massive MIMO solutions.
Table of Contents
1. Introduction to Massive MIMO 10
2. Requirements and solutions

in an evolving 5G network 20
3. Massive MIMO solutions 42
4. Guiding principles for how to select

a radio solution 72
5. Future developments 86
6. Summary 100
7. Abbreviations 102
8. References 106
1. Introduction to
Massive MIMO
1. Introduction to Massive MIMO

Ericsson | The Massive MIMO handbook 1. Introduction to Massive MIMO 11
Outline
Outline Ch
No
• What Massive MIMO is
• Why Massive MIMO is used
• How Massive MIMO works
The main purpose of this introduction is to briefly explain what To deliver performance enhancements, Massive MIMO makes
Massive MIMO is, why it is used and how it works. use of multi-antenna technologies and antenna array properties.
The multi-antenna technology components, viz. beamforming,
This chapter starts with a brief history of Massive MIMO, followed MIMO and nullforming, are briefly introduced and their relevance
by a definition of Massive MIMO and a discussion of the relation in the network is also discussed. The properties of antenna
to conventional solutions. The main reason why Massive MIMO is arrays and how they can influence network performance are also
so relevant in today’s 5G mobile networks is the ability to enhance introduced.
coverage, capacity and user throughput, in a cost efficient
manner, which is then shortly explained. This brief introduction gives a short glimpse of the very rich and
fascinating topic of how Massive MIMO works. Interested readers
can however find a more thorough description in [TP1-9] and in
[1].

Massive MIMO is here and now E

Massive MIMOahead
Speeding is herewith
and 5G
now Q
Speeding ahead with 5G
5G roll out progress 336 165
30 074
Mar. '19 Apr. '19 May. '19 Jun. '19 Jul. '19 Aug. '19 Sep. '19 Oct. '19 Nov. '19 Dec '19 Jan '20
SKT
CSP 1 KT
CSP 2 LGU+
CSP 3 Total
Massive
AAS vsMIMO vsGHz
Classic 3.5 RRU in 3.5
radios GHz for
in Ericsson areaEricsson area
100%
12% 17% Massive MIMO
80% 37%
Remote radio
60%
40% 88% 83%

63%
20%
0%
CSP A1
MNO
CCD bare
MNO
metal
CSP 2B CSP
MNO3C
Multi-antenna technologies, such as beamforming and spatial The use of Massive MIMO is growing fast with the deployment of
multiplexing, have been known for many decades. The mobile 5G networks. Massive MIMO is used as an intrinsic component in
industry began to explore these technologies almost 30 years all 5G networks from the start and the fraction of Massive MIMO
ago. For 3G/WCDMA, these solutions never reached significant solutions deployed in the networks is high. 5G networks are
commercial success, and the number of deployed solutions were built in a similar way across the world. The build-out starts from
few. For 4G/LTE, beamforming and MIMO solutions were more the dense urban areas where the requirements are highest and
advanced and more efficient and have been used commercially continues gradually outwards to areas with lower traffic.
in many deployments. Massive MIMO solutions, the latest
evolution of the multi-antenna technologies, have been deployed The graph shows the rollout of radio sites in a country in East
for 4G since 2017/18. With the deployment of 5G/NR, Massive Asia with the accumulated number of radio sites for the three
MIMO is an inherent technology component from the start. service providers (also called communication service providers
There is now a significant uptake of 5G Massive MIMO solutions (CSP), referred to in the figure as CSP 1-3). The lower graph
globally, and it is expected that the Massive MIMO share of the shows the fraction of Massive MIMO versus Remote Radio
network will continue to grow. solutions for the three CSPs, respectively. The fraction of Massive
MIMO is high in all three networks, ranging from 63% to 88%.
The ratio of Massive MIMO products versus Remote Radio is
usually different for different deployment environments, higher
for dense urban areas and lower for suburban areas. For a given
deployment environment, it is expected that the ratio of Massive
MIMO compared to Remote Radio will increase over time. Over
two years, the number of 5G base stations increased by a factor
of 10 from a base of 30k in March 2019 to more than 300k in
March 21. Massive MIMO radios were the radio of choice for the
mid-band spectrum at 3.5 GHz with about 80% Massive MIMO
versus 20% conventional Remote Radio solutions in Ericsson
areas.
Note that in this example, the deployment is made in an

area which is predominantly high-traffic and therefore, the
percentage of Massive MIMO is relatively high. In other types
of deployments, this percentage may be lower.

Massive MIMO solution
Massive MIMO solution
+ =
Baseband Massive MIMO radio Massive MIMO

solution
Conventional solution
+ + =
Baseband Remote radio Passive Conventional

antenna solution
There are many definitions of Massive MIMO. In this handbook, There are no hard boundaries between Massive MIMO solutions
the following definitions are used: and conventional solutions. A conventional solution, or a non-
Massive MIMO solution, typically comprises a passive antenna
• Massive MIMO is a concept where multi-antenna techniques and a remote radio unit using a low number of radio chains**
exploit massively many antennas with dynamically adaptable (typically 2, 4 or 8). A Massive MIMO solution is typically much
input and/or output signals. more capable than a conventional solution in terms of executing
the multi-antenna features.
• A Massive MIMO solution is an implementation of Massive
MIMO, consisting of both hardware (Massive MIMO radio) and *) In the Ericsson product portfolio, the baseband solution
software parts (Massive MIMO features). is realized with different form factors of hardware to better
• A Massive MIMO radio is a hardware unit that comprises an fit service providers’ deployment architectures (e.g. outdoor
antenna array, a large number of radio transmitter and receiver deployment, indoor rack mounted deployment) and will
chains (in this document, large is assumed to be ‘significantly have different naming to reflect the supporting radio access
more than 8’), and parts of the baseband functionality, all technology (e.g. baseband support 2G/3G/4G/5G, RAN
tightly integrated, see [Ch. 3.1, p. 45-46]. processor support 4G5G).
• A Massive MIMO feature is a multi-antenna (software) feature, **) A transmitter radio chain (T) contains the functionality
such as beamforming, nullforming or spatial multiplexing, or a required to transform a baseband signal to a high-power radio
combination of these. Massive MIMO features can be executed frequency signal that can be transmitted over the air and
in the Massive MIMO radio or in the baseband* or in both conversely a receiver radio chain (R) contains functionality to
locations [Ch. 3.1, p. 45-46]. transform a radio frequency signal to a baseband signal. The
number of transmitter chains does not need to be the same as the
There are many concepts that are similar or identical to Massive number of receiver chains. In this document, it will be assumed
MIMO. Examples include advanced antenna systems or active that these numbers are equal and when referring to ‘radio
antenna systems, both abbreviated as AAS. It should also be chains’, both the transmitter and the receiver parts are included.
noted that the definition of Massive MIMO within academia is The number of radio chains impacts the beamforming and MIMO
somewhat different to that used by industry. capabilities. Also, it affects the requirements on signal processing
and the total complexity and cost of the product.

Massive MIMO is addressing key

network characteristics
Massive MIMO is addressing key network
characteristics
Massive MIMO network Conventional network
Total cost of ownership

User throughput User throughput
per capacity
Increased Reduced
user experience TCO
Increased user
experience
Increased
Increased network
coverage capacity
User Coverage Distance

User experience Capacity Served traffic
experience requirement Remote radio Massive MIMO
requirement
Increased capacity, user throughput and coverage in both uplink and downlink for existing
and new frequency bands. Reduced Total-Cost-of-Ownership (TCO) per capacity
Massive MIMO can increase network coverage, capacity, The middle figure shows user throughput as a function of served
and user throughput using multi-antenna technologies such traffic. When the served traffic stays the same, the benefit will
as beamforming, nullforming and spatial multiplexing (MIMO). be seen in terms of increased user throughputs also in a loaded
These exploit spatial properties of the antenna array and the network. At the same time, the figures also show that Massive
propagation channel. These techniques are briefly described in MIMO can serve more traffic for a given data rate requirement.
the following three pages, [Ch. 1, p. 15-17], and in more depth This means that the network is prepared to handle future traffic
in [TP2], [TP3], [1, Ch. 6] and [2]. growth.
The figure on the left illustrates single user throughput as a Massive MIMO can also be more cost efficient in some scenarios.
function of distance to the serving base station in an unloaded In the figure on the right, different colors describe different cost
network for a conventional network and a Massive MIMO contributions. Here, cost efficiency is measured as total cost of
network. It is assumed that both networks have the same system ownership (TCO) per delivered capacity and Massive MIMO can
settings, e.g. bandwidth and output power. Massive MIMO be more cost efficient in this metric when the load is high or
increases the coverage of all throughput levels. Specifically, it will be high during the investment period. This will be further
increases the user throughput for most distances (the peak rate explained on [Ch. 2, p. 39] and in [TP9.4].
is assumed to be the same for both networks). These aspects are
thoroughly elaborated in [1, Ch. 6, 13].

Massive MIMO offers a diverse set

of techniques addressing the different
network needs No
Beamforming MIMO – Spatial multiplexing

Increase signal quality by spatially focusing power Increase data rates via parallel data streams
High beamforming gain Low beamforming gain Single-User MIMO Multi-User MIMO
1
Layer 1 Layer
Layer 2 Layer 2
Layer 3
Nullforming Layer 3
Reduce interference
by avoiding “directions”
There are basically three multi-antenna technology components Nullforming: Nullforming is a form of beamforming that strives
that contribute to the increased performance of Massive MIMO, to lower the beam gain in certain directions or reduce it to
viz. beamforming, nullforming and spatial multiplexing (also zero. By intentionally designing the beam shape to have nulls
referred to as MIMO). These technologies are applicable to both or lower gain in the directions where the victim transceivers
downlink and uplink and are briefly introduced below. are, interfering signals can be reduced. The picture illustrates
nullforming that mitigates inter-cell interference while
Beamforming (BF): The purpose of beamforming is to amplify communicating with a UE of interest in the own cell. Note:
transmitted/received signals more in some directions than Nullforming is usually a concept used for the downlink. For the
others. A common goal is to achieve a high BF gain in the uplink, the concept corresponding to nullforming is interference
direction of the device of interest to improve link quality in terms rejection combining (IRC).
of signal-to-interference-and-noise-ratio (SINR). Improved link
quality translates into better network coverage, capacity, and Spatial multiplexing: In massive MIMO, it is possible to
user throughput. As the energy conservation principle applies, multiplex several data streams on the same time-frequency
having a higher gain in some directions necessarily implies lower symbol. This is often called spectrum re-use and the data streams
gain in others. Thus, a high gain beam is often narrow while a are commonly referred to as layers. The multiple layers can all
low gain beam has a more isotropic radiation pattern. Note: go to the same device or split to different devices. The case when
The beam is often depicted as one lobe directed to the receiver. all the layers belong to the same transmitting/receiving device
In reality, the beam distributes power in many directions. This is is referred to as single-user MIMO (SU-MIMO) while the case
further described in [TP3] and in [1, Ch. 4-6]. involving spatial multiplexing of multiple transmitting and/or
receiving devices is called multi-user MIMO (MU-MIMO). Spatial
multiplexing can increase user throughput and network capacity.
These fundamental technology components can be combined

in various ways to form complete Massive MIMO features, as is
shown in [Ch. 3.3] and [TP7]. More details on the multi-antenna
technology components can be found in [TP2], [TP3] and [1, Ch.
6] and examples on performance in [TP9].

The fundamental massive MIMO techniques

have different levels of importance Box lines
ZOOMA
The gree
things in
Can we c
Beamforming MIMO – Spatial multiplexing
Increase signal quality by spatially focusing power Increase data rates via parallel data streams
High beamforming gain Low beamforming gain
Single-User MIMO Multi-User MIMO
1
Layer 1 Layer
Layer 2 Layer 2
Layer 3 Layer 3
Nullforming
Reduce interference
by avoiding “directions”
Main performance tools
Further performance boosters
Main performance tools Further performance boosters

Beamforming and single-user MIMO are very central to Massive MU-MIMO improves performance at high traffic loads and
MIMO. The ability of beamforming to increase the received good channel conditions. These are conflicting requirements
signal level is crucial to obtain high performance. The impact as high traffic loads often lead to higher inter-cell interference
on interference levels may however be small or non-existent, levels, which means worse channel conditions. Not all cells
unless intentional nullforming is used. This is because the benefit significantly from MU-MIMO. Gains from beamforming
total transmitted power is approximately the same, either are of greater importance. Compared to SU-MIMO, there are
beamforming or single-user MIMO is used or not and hence the many more requirements on MU-MIMO to reach meaningful
average interference levels are similar. Substantial beamforming performance improvements. MU-MIMO is nevertheless a great
gains can be achieved in a wide range of situations, regardless capacity enhancement tool for highly loaded cells as is also
of downlink/uplink, traffic load, or if the user is in a good or bad further discussed in [TP3], [TP9] and [1, Ch. 6].
spot. Capacity and user throughput generally improve. Improved
coverage is an obvious and especially important benefit, Intentional nullforming to selected users serves to reduce
facilitating reuse of existing site grids for higher frequencies. interference to those users. It is a key sub-component of
multi-user MIMO to mitigate intra-cell interference and it is also
Spatial multiplexing via single-user MIMO (SU-MIMO) benefits commonly used on the receiver side in both uplink and downlink
from high signal levels. Beamforming helps to improve signal to suppress inter-cell interference. The inter-cell interference may
levels which then can be exploited for single-user spatial be substantial in some scenarios and then gains from nullforming
multiplexing. Particularly in downlink, more than one layer to a on top of the performance given by beamforming and single-user
specific user can often be supported in rather large parts of a cell. MIMO can be significant, which is addressed by interference
This contributes to its relatively general applicability. sensing, see [TP7.3] and [TP9].

A Massive MIMO radio uses antenna arrays

to enable Massive MIMO techniques No
Small array Conventional system Massive MIMO

Two radios Large array Large array
Low gain, fixed beam Few radios Many radios
High gain, limited Multiple steerable
beam flexibility and shapeable beams
The capabilities of doing beamforming and MIMO depend on he properties of the antenna array can be shaped by
T
the ability to control the amplitude and phase of parts, or even partitioning the array into an array of subarrays (AOSA), where
single elements, of the antenna array. The dimensions and the each subarray is controlled individually. A subarray is a group of
properties of the antenna array affect the network performance element pairs and is illustrated with red rectangles in the figure
and are key inputs to the product portfolio. The antenna array, above. The array in the middle figure has two subarrays stacked
its structure and its capabilities are here shortly introduced. For horizontally in a 1x2 AOSA and the array in the right figure
an extensive description of antenna array properties, see [TP1], consists of 32 subarrays stacked horizontally and vertically in
[TP2] and [1, Ch. 3-4]. a 4x8 AOSA structure. The beam pattern of the antenna array
is controlled by steering the amplitude and phase of the signals
The maximum antenna gain is proportional to the total antenna fed to the subarray. The purpose of using subarrays is to get
array area. A small antenna provides low gain, while a large the appropriate steerability of the array while keeping the
antenna provides high gain. See the figures on the left and in complexity, and hence the cost, at an adequate level. The finer
the middle. It also follows that the width of the main lobe, in partition (smaller subarrays), the better steerability, but higher
horizontal/vertical dimension, is inversely proportional to the cost and complexity, since more radio chains are needed for the
size of the antenna array in horizontal/vertical dimension. same size of the array.
A wide antenna will, therefore typically produce narrow lobes
and a narrow antenna will produce wide lobes. The size and shape of the antenna, and how it is partitioned into
subarrays, affect Massive MIMO performance. Design choices
for these parameters are important to provide products with the
desired properties. Also of particular importance is the effective
isotropic radiated power (EIRP), which is a combination of
output power and antenna gain.

Massive MIMO — future-proof

solution providing superior experience
No chan
Mbps Downlink user throughput illustration
750
500
250 FDD bands FDD + NR TDD Conventional
100
Downlink user throughput

50
Traffic demand according to 2025
10
predictions
FDD + NR TDD Massive MIMO
By applying the methods outlined in previous slides, Massive In this time period, typical frequency division duplex (FDD)
MIMO can support performance that is superior to that of bands cannot fully support the traffic demand. Adding time
conventional radios. The figure illustrates downlink user division duplex (TDD) mid-band spectrum, such as the C-band
throughput for an example deployment scenario given the in the US, with conventional remote radios improves the
expected traffic demand of 2025 and compares different performance significantly, but there are still many users with
solutions. a rather poor throughput, meaning that the network is likely
operating close to its capacity limit. The combination of TDD
The deployment scenario is Atlanta midtown (US dense urban mid-band spectrum and Massive MIMO, on the other hand,
scenario), and results are based on multi-band simulations for provides superior user experiences and a network operating well
three different radio deployment options: below its capacity limit. This shows that adding NR TDD mid-
band with Massive MIMO provides a future-proof investment
• FDD bands only: FDD 700 MHz with 20 MHz bandwidth and
that can handle the anticipated traffic growth for many years.
2T2R radio, and FDD 2 GHz with 40 MHz bandwidth and 4T4R
A much more detailed analysis is given in [TP9].
radio

• FDD bands plus 3.5 GHz TDD with 100 MHz and 8T8R radios
• FDD bands plus 3.5 GHz TDD with 100 MHz and Massive MIMO
(64T64R)

Summary — Massive MIMO improves

relevant mobile network requirements
N
● Massive MIMO has reached general use

and is growing rapidly on the market
● Massive MIMO solutions consist of a hardware
and a software part
● Massive MIMO can provide better coverage,
improved capacity, increased user throughput
and better cost efficiency
● Massive MIMO features exploit the spatial
domain and build on techniques such as
beamforming, nullforming and spatial
multiplexing
Massive MIMO makes use of many antennas with adaptive Massive MIMO makes use of multi-antenna features and the
input and output signals. A Massive MIMO solution consists of antenna array characteristics to reach the desired performance
a radio hardware part (Massive MIMO Radio) and multi-antenna benefits. These characteristics are exploited by the Massive
features (Massive MIMO features). MIMO hardware and software solutions. The most important
multi-antenna features are beamforming and single-user MIMO.
Massive MIMO techniques are applicable in both downlink These features provide benefits in almost all environments.
and uplink and can provide better coverage, capacity and user Examples of other multi-antenna features are multi-user MIMO,
throughput, which are useful characteristics in a mobile network that can provide significant benefits in some environments
with continuously growing traffic. Massive MIMO solutions can under certain conditions, and nullforming that can suppress
also be more cost efficient than conventional Remote Radio interference to and from certain identified users or suppress
solutions in many scenarios. inter-cell interference. The antenna array characteristics are
also important to shape the properties of a Massive MIMO
solution, e.g. size, number of elements in horizontal and vertical
dimensions and subarray partitioning.

2. Requirements and
solutions in an evolving
5G network
2. Requirements and solutions in an evolving 5G network

Ericsson | The Massive MIMO handbook 2. Requirements and solutions in an evolving 5G network 21
Outline
No
• Network requirements
• Network solutions
• Network variations – each site is unique
● Spectrum properties affecting the radio

solution
• Deployment constraints
The purpose of this chapter is to discuss the most important he requirements for specific sites are also discussed as they
T
network performance requirements and constraints impacting depend on the environment in which the sites are deployed,
5G network evolutions and highlighting where Massive MIMO the actual load and the load variations over time.
can make a difference.
I n addition to the network requirements, there are also
G networks need to be evolved to handle increased traffic
5 constraints to consider, viz. ease-of-deployment, cost efficiency
growth and the expectations on enhanced user experience, and energy performance. These will also be discussed.
both for old and new services, which leads to requirements on
increased coverage, capacity and user throughput. The possible
ways to meet these requirements are outlined and the specific
properties and preferred use of TDD mid-band, FDD mid-band
and mmWave, respectively, are discussed.

A new 5G experience
A new 5G experience No chan
Example 1: Extended reality services Example 2: Industry automation services
There is currently a strong push in consumer and enterprise Many of the new services will need higher user throughput, lower
communities for new mobile services that reach beyond evolved latency and improved reliability compared to today’s best-effort
mobile broadband (MBB). Examples of such services include MBB services. Lower latency and improved reliability can be
extended reality (XR), advanced gaming and industrial control translated into higher capacity demand as these requirements
applications. Fixed Wireless Access (FWA) is another service that imply a higher consumption of radio resources than best-effort
already has proven to be a big success, [3]. See [Ch. 5.1] MBB to keep required margins.
for further discussions on new services.
To meet the requirements of new services, the entire telecom
There are several reasons for the recent interest in new services. ecosystem needs to be evolved. Several new use cases, and
Society is expecting a leap forward in the service domain in the functionality required to support them, are developed in
the coming years and the industry needs to find new business standardization forums. The standardization bodies provide
revenue streams. The vast performance improvements offered an extensive toolbox that network and UE providers need to
by 5G and new mid- and high-band spectrum can make these choose between; what features to implement and when to do
new services a reality. it. Implementing new network features often requires addition
of software and hardware, and although the standards specify
much of the functionality, there is still a possibility to differentiate
beyond the basic standard support using different products. For
example, as described in [Ch. 1] and later in this chapter [Ch. 2,
p. 31] and more extensively in [TP3] and [TP9], Massive MIMO
is one solution that can offer capabilities and performance that
exceed conventional solutions.

Total global mobile data traffic – excluding forecast period, particularly in the uplink. improved device capabilities; an increase
Ericsson | The Massive MIMO handbook traffic generated by Fixed Wireless Access 2. Requirements
At the end of 2023, video traffic isand solutionsinindata-intensive
an evolving 5G network
content; and growth 23
(FWA) – is expected to reach 130 EB per estimated to account for 73 percent in data consumption, due to continued
month at the end of 2023. This is projected of all mobile data traffic. improvements in the performance of
to grow by a factor of around 3 and to reach Populous markets that launch 5G early deployed networks.
403 EB per month in 2029. When FWA is are likely to lead in terms of traffic growth An example of these differences
included, total mobile network traffic is over the forecast period. 5G’s share of can be seen in the contrast between
anticipated to reach around 160 EB mobile data traffic is estimated to be the Sub-Saharan Africa region, where
25 percent at the end of 2023, an increase
Higher capacity is required in networks

per month at the end of 2023, rising to average monthly mobile data usage
an expected 563 EB per month by the from 15 percent at the end of 2022. This share per smartphone is expected to be
end of 2029. Predicted traffic growth is forecast to grow to 76 percent in 2029. 6.7 GB at the end of 2023, and the
up to 2029 includes an assumption that Gulf Cooperation Council (GCC)
to deliver a 5G experience
an initial uptake of XR-type services, Traffic growth varies across regions countries, where average monthly usage
including AR, VR and mixed reality (MR), Traffic growth between years can be is expected to be 30 GB per smartphone
will happen in the latter part of the highly volatile and can vary significantly at the end of 2023. The global monthly
forecast period. However, if adoption is between countries, depending on local average usage per smartphone is
stronger than expected, data traffic could market dynamics. Globally, the growth anticipated to reach 21 GB in 2023 and is No
increase significantly more than currently in mobile data traffic per smartphone forecast to reach 56 GB by the end of 2029.
anticipated toward the end of the can be attributed to three main drivers: The average monthly mobile data
usage per smartphone in North America
is expected to reach 66 GB in 2029,
Figure 9: Global mobile network data traffic (EB per month) as unlimited data plans and improved
5G network coverage and capacity
Global mobile data traffic (EB per month)
FWA (3G/4G/5G) Mobile data (5G) Mobile data (2G/3G/4G) increasingly attract new 5G subscribers.
MBB data traffic 600
Data traffic will increase significantly
in line with the expected uptake of
alone is growing by FWA (3G/4G/5G)
550 gaming, XR and video-based apps.
50-70%
These experiencesMobile datahigher
require (5G) video
500 resolutions, increased uplink
Mobile data traffic,
(2G/3G/4G)
and more data from devices off-loaded
450 to cloud computing resources to satisfy
per year
users. FWA has started to affect overall
400
traffic patterns, accounting for a large
share of total network data traffic.
350
300
250 Globally, average monthly mobile

data usage per smartphone is
200 expected to reach 56 GB, rising
from 21 GB at the end of 2023.
56
150
100
50
0
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
GB
Globally, traffic in mobile broadband (MBB)* networks has * MBB, often defined as ‘best effort’ data traffic, is the traffic type
increased 50-70% each year for many years, see figure**. This that dominates in mobile networks today. MBB networks are
growth is expected to continue at a similar or higher pace in the primarily built and expanded to meet the continually growing
future. To cater for this traffic growth, mobile networks must data consumption need. Mobile broadband services include web
support a corresponding capacity growth. As discussed on the browsing, social networking, video streaming (Netflix, YouTube),
previous page, the introduction of new services like extended email and messaging, online gaming, audio services, and file
reality (XR) could further accelerate the rate of traffic increase. sharing.
Evolving a network requires increased capacity over time to ** The source of the graph is Ericsson Mobility Report November,
maintain high service quality. [4].

User experience challenged by congestion

in busy
User areas during
experience busy
challenged byhours
congestion Title ch
Figure
in busy areas during busy hours +y
Time to
moved
Time-to-content per coverage area
based on 800 million user activities measured 24/7 in week 45, 2021
European service provider A European service provider B
10 10
8
Time-to-content (s)
Time-to-content (s)
Busy times worst-case Busy times worst-case
6 6
Average Average
4 4
2 2
All coverage areas ranked on worst-case time-to-content All coverage areas ranked on worst-case time-to-content
User experience: Great Good Fair Poor
Throughput “at click” 20 Mbps 10 Mbps 5 Mbps

Time-to-content 1.5 s 2.5 s 4.0 s
There is an important relation between user experience and The figures show time-to-content as a function of all cells sorted
network capacity. A network that is dimensioned to handle from worst to best busy-hour time-to-content for two different
the peak load can consistently deliver a high user experience, networks. Each cell in the network is measured every 15 minutes
whereas networks that are insufficiently dimensioned will over a period of one week, and the darker colors represent the
experience impacts on user performance during high load. median (50th percentile) and the lighter colors represent the
busy-hour (90th percentile) time-to-content measurements.
Time-to-content is an important metric that describes the time In total there are 800 million measurements.
it takes to access a service after initialization, for example, the
time it takes to load a web page after clicking the link. This Service provider A (left figure) has a high-capacity network
metric is what really matters for the end users. Studies done by that is rated best-in-class. The network can therefore offer high
Ericsson show a clear correlation between time-to-content and performance in all cells at any time. Busy-hour time-to-content is
end-user experience. This is illustrated by the coloring explained rated good or great (<2.5 seconds) for most of the cells and poor
in the lower legend, where user experience is rated from poor to performance for only about 4% of the cells. Service provider B
great. Furthermore, a consistently low time-to-content is crucial (right figure) has a lower capacity network and time-to-content
for online businesses; be it e-commerce, e-travel, e-banking, or is rated Poor (>4 seconds) for about 30% of the cells.
other.
To summarize, it is very important to have sufficiently high
capacity in all cells to avoid congestion during peak hours.
Congestion leads to degraded user experience and poor
time-to-content.

Network solutions to increase capacity
y
nc
cie
Site density
ffi
le
tra
ec
Sp
Spectrum
Higher capacity can be achieved by increasing spectrum, spectral efficiency, and site density
Massive MIMO systems increase spectral efficiency (green arrow) compared to conventional systems
There are several ways to deliver more capacity in a mobile pectral efficiency, i.e. the number of bits that can be transferred
S
network. The preferred way of doing this depends on several on a specific spectrum bandwidth, measured in [bps/Hz], can be
factors, e.g. availability of spectrum, cost efficiency, capital improved in many ways. A very efficient way of doing this is by
expenditures (CAPEX) and operational expenditures (OPEX) exploiting the spatial domain with multi-antenna features such
budgets, accessibility and cost for new sites, etc. as beamforming, MIMO, and nullforming. Thus, Massive MIMO
provides important tools to further increase network capacity.
In the figure, capacity is represented by the volume of the red The effectiveness of these features varies, as indicated in [Ch. 1,
box, and there are essentially three ways of increasing capacity, p. 16] and elaborated in [TP9] and [1; Ch. 13].
adding spectrum, increasing spectral efficiency in the network
and increasing the site density. Increasing the site density in the network is a third way of
increasing capacity. There are different ways to densify the
Adding spectrum on new or existing frequency bands is a very network, for example, sites on roof-top level, which is often
efficient way to add capacity. Many loaded sites use all available preferred, or sites on street level. Radio sites are, however, often
spectrum already. With the introduction of 5G, however, expensive to acquire and maintain. Hence, site densification is
most countries have released new spectrum, in most cases on usually exploited when the other options are exhausted.
mid-band around 3.5 GHz or high-band (mmWave) around
28-39 GHz. The amount of new 5G spectrum is considerable,
typically in the order of 100 MHz or more, and can hence
provide much capacity. To increase capacity in a 5G network, it
is, therefore, imperative to make efficient use of the available

spectrum.

Spectrum is key to unlocking

substantial
Spectrum is keycapacity
to unlocking substantial
capacity
Bandwidth/
throughput
High-band
(mmWave)
Mid-band TDD
Mid-band FDD
Low-band FDD
Coverage
<1 GHz 1.4-2.7 GHz 2.3-5 GHz 24+ GHz

3GPP
spectrum
10 9 1010
Existing frequency bands New 5G frequency bands
Spectrum is one of the most valuable radio network assets. Existing networks are typically built to provide coverage
Deploying and being able to fully utilize the new 5G frequency ‘essentially everywhere’ using traditional frequency bands in the
bands will be key to unlock substantial capacity and provide a mid-band FDD region, e.g. 1.8 GHz (GSM) or 2.1 GHz (3G). For
superior user experience. higher frequency bands, coverage using the existing site grid will
be more challenging [1; Ch. 3], and there are two main reasons
Many countries have released substantial amounts of new for this. Firstly, radio propagation becomes more challenging the
spectrum for 5G deployments in the mid-band region covering higher the frequency (the physical reality). Secondly, 5G services
~3.5-5 GHz and high-band (mmWave) region covering often have higher throughput requirements; that is, 5G networks
~24-40+ GHz. Two key characteristics of these new frequency need to be dimensioned for higher data rates compared to
bands compared to existing frequency bands are higher carrier 2G-4G networks. The latter is often referred to as data coverage
frequencies and large bandwidths. (or sometimes App coverage) as opposed to access coverage,
which relates to the ability to connect to the network and have
sufficient coverage for low-rate signaling and control channels.

Massive MIMO to maximize the spectrum

usage on existing site grid
Bandwidth/
throughput
High-band
(mmWave) Beamforming (BF)
BF
Mid-band TDD
Frequency
interworking
Mid-band FDD BF
Low-band FDD
Coverage
The large bandwidths available on the new and higher A careful network planning and interworking between frequency
frequency bands have the potential to offer substantial network bands will also be essential to ensure that the strengths of
improvements, but to fully reap the benefits in terms of coverage, each frequency band are explored. As an example, capacity
capacity, and user throughput, will require service providers to only needs to be offered where it is needed, meaning that it
use the available toolbox to its fullest extent. might not be necessary to strive for ubiquitous coverage on all
frequency bands. For mmWave, the cost of building the network
The most important aspect when deploying new and higher for ubiquitous coverage might overshadow the return in terms of
frequency bands is to make the spectrum accessible (where increased performance. See [TP9] for further details.
needed) in the network. Beamforming offered by Massive MIMO
products is key for improving downlink and uplink coverage and assive MIMO can also substantially increase the spectral
M
thereby maximizing the region where users get access to the efficiency and user throughput by using techniques such as
superior performance offered by the vast bandwidth at higher spatial multiplexing (both SU-MIMO and MU-MIMO) and
frequency bands. nullforming.

Increasing site density improves both

coverage and capacity but comes Minor cha
Removed
with a cost
Densifying the network by adding new sites will improve both Densification comes with some deployment and performance
coverage and capacity. The main drawbacks of densification are challenges that Massive MIMO can help alleviating. One
cost and ease-of-deployment considerations. The cost and time challenge is to find good site positions that improve network
required for site acquisition can be substantial due to finding performance by covering areas with high traffic and also
sites, negotiating with building owners, site preparation in qualify from an ease-of-deployment perspective. Densification
terms of construction, regulatory approvals, etc. The operational may also increase the inter-site interference, in particular in
expense (OPEX) per site can also be significant due to site rental, heterogeneous deployments where nodes of different types,
transmission and regular site maintenance of the equipment. e.g. Macro and Street, operate at the same carrier frequency.
Consequently, densification may impact the network’s cell
Due to the high cost, network site densification has not yet plan, that could result in a new tedious and costly cell planning
taken off on a large scale, and is currently used mostly locally exercise. Massive MIMO offers tools such as cell shaping that
to address areas with very high traffic. There are, however, can be used for automated cell planning based on, for example,
indications that site densification might be inevitable to support artificial intelligence and machine learning, hence simplifying the
the network demands of the future. New services like XR that site densification process [TP7.3].
have high requirements, in particular on uplink coverage and
capacity, might drive a more large-scale site densification. Cost
and ease-of-deployment aspects of densification have also
triggered a renewed interest in small cells or street radios that
have very small form factors and are easier to deploy.

Requirement variations in the network
Requirement variations in the network No

Als
Mi
red
Different deployment Traffic load variation Traffic load variation

environments for sites over time
MBB busy hour traffic per site
Relative traffic volume

Traffic [MB]
0%
0% 50% 100%
Dense urban Urban Suburban Rural
ISD 100-500m 600-1000m 1500-4000m ~5-20km High Medium Low Time [hours]
Coverage ‘essentially everywhere’ and capacity where needed

Different requirements in different parts of the network
Each site in the network is unique, and both network the left (RED) to the medium loaded sites in the middle (BLUE)
performance requirements and constraints can vary significantly to the low loaded site to the right (GREEN). The value on the
between different sites and different parts of the network. This vertical axis is the level of traffic load in the site. This traffic
will affect the choice of radio solutions. load distribution is a function of the distributions of sites, users
and the use of services in different parts of the network. It is
Different deployment environments have different therefore very similar for 3G and 4G and is expected to remain
characteristics in terms of building height distribution, inter-site roughly similar also for 5G. The distribution looks similar in most
distances (ISD), subscriber density, use of services, building networks. The high, medium and low loaded sites can be found
wall thickness, window glass characteristics, occurrence in any part of the network. The performance requirements will
of vegetation, etc. The characteristics of the deployment be different as they depend on the actual traffic load on the
environments have an impact on which products serve future site. This load distribution is an important input to portfolio
network requirements in the most cost efficient manner. The segmentation. Also, it impacts the product selection of the
spatial distribution of users is especially important for Massive sites. Later in this chapter, different load segments, high (RED),
MIMO. The area where the Massive MIMO solution works medium (BLUE), low (GREEN), will be defined which will be
efficiently in terms of providing coverage, user throughput useful when selecting products for the sites, respectively.
and capacity is often referred to as the service area or angular
coverage area, and it is important that the service area covers There are also significant load variations over time in the cells, as
the area where users are located. illustrated in the figure to the right. During peak hours, the load
can be considerably higher than during off-peak. If the capacity
Traffic load varies between sites in real networks. A few sites in the cell is not sufficient, the cell will be congested and there
carry a significant amount of the total network traffic, whereas will be an impact to the end user experience, as shown in [Ch. 2,
most sites are modestly loaded. The data in the graph in the p. 24].
middle figure show a typical load distribution in a 4G network.
In this graph, the sites are ordered from the high loaded site to

Deployment environment impacts

the desired Massive MIMO properties
Deployment environment impacts the
desired Massive MIMO properties
Vertical domain beamforming—high-rise deployment
(2x1) subarrays
ISD Dense urban Urban Suburban Rural

Inter-site distance 100-500m 600-1000m 1500-4000m ~5-20km
Vertical domain beamforming—low-rise deployment

Horizontal (4x1) subarrays
domain
beamforming
Networks must provide coverage in the entire service area (the The lower left figure illustrates beamforming in the horizontal
area where users are located). For most macro networks, the domain, and in most cases a wide horizontal coverage area is
service area typically includes the entire horizontal domain, needed as users are often spread homogeneously in the cell.
while which part of the vertical domain that must be covered This also means that horizontal domain beamforming with as
is highly scenario dependent, ranging from a small range of many antenna columns as possible is typically very effective for
vertical angles in suburban/rural environments to potentially a improving the performance.
large range of vertical angles in dense high-rise environments.
This observation has an impact on the choice of Massive MIMO The right figures illustrate beamforming in the vertical domain.
product, where smaller subarrays are needed to have a wide The high-rise environment (upper right figure) with a large
vertical coverage area, leading to many radio chains in order to spread of users in the vertical domain requires a large vertical
maximize antenna area, hence total antenna gain, while larger service area, which means that smaller subarrays are needed.
subarrays suffice for a narrower vertical coverage area. In the example, 2x1 subarrays are used, meaning that the
8x8 antenna array is partitioned into a 4x8 array of subarrays
The figures illustrate the required service area and the (AOSA) structure. The large spread of users in the vertical
effectiveness of beamforming. The required service areas domain also means that vertical domain beamforming is useful,
are illustrated by the red dashed lines originating from the and the 4x8 AOSA structure facilitates efficient vertical domain
antenna. It is important that the coverage area of the antenna beamforming. The low-rise environment (lower right figure)
configurations, that is, the area where UE specific beamforming with a small angular spread of users in the vertical domain
can be done without a significant gain drop, matches the requires a smaller vertical service area, which means that larger
required service area. subarrays can be used. In the example, 4x1 subarrays are
used, meaning that the 8x8 antenna array is partitioned into a
2x8 AOSA structure. A (2x1)(4x8) antenna structure, i.e. a 4x8
AOSA with 2x1 subarrays, as used in the high-rise environment,
would in this environment essentially have degrees-of-freedom
(beamforming directions) of no or little use. In other words, the
performance of a 64T (2x1)(4x8) antenna structure and a 32T
(4x1)(2x8) are similar in the low-rise environment as the benefit
from vertical domain beamforming is limited.

Performance example: How deployment

environment affects the Massive MIMO
In
efficiency
1 2 3
100m
Radio ISD Radio Radio ISD Radio Radio ISD Radio

50m 30m 20m
250m 500m 750m
9m
Relative capacity gains Relative capacity gains Relative capacity gains

@10 Mbps cell-edge throughput target @10 Mbps cell-edge throughput target @10 Mbps cell-edge throughput target
Most Massive MIMO Most Massive MIMO

gains achieved already gains achieved already
with 32T32R with 32T32R
As discussed in [Ch. 1, p. 14], Massive MIMO can substantially In a dense urban high-rise deployment (figure 1), there are
improve both the uplink and downlink spectral efficiency both substantial gains from Massive MIMO compared to remote
for single users (bps/Hz) and for the network (e.g. bps/cell/ radios, and 64T64R provides significant gains over 32T32R.
Hz), hence ensuring an efficient use of the available spectrum. Dense urban high-rise is a deployment where a capable Massive
The gains from Massive MIMO depend on aspects such as the MIMO radio product with advanced beamforming features
antenna array structure, algorithm choices and the deployment excels. Due to significant spread of users in the vertical domain,
environment. This is briefly illustrated in the following example both horizontal beamforming and vertical domain beamforming
and further described in [TP9]. are useful. Note: In some high-rise deployments, indoor systems
are often installed in the high-rise buildings. This reduces much
The figures show downlink capacity gains of four different of the need to cover these buildings from the outdoor radio sites
products in three different deployment scenarios, dense urban and the benefit of vertical beamforming is in practice lower in
high-rise (figure 1), urban mid-rise (figure 2) and suburban those cases.
(figure 3). The products are 4T remote radio, 8T remote radio,
32T Massive MIMO and 64T Massive MIMO, and gains are In the urban mid-rise and suburban deployments (figures 2 and
relative to the 4T remote radio product. For these products 32T 3), there are still substantial gains from Massive MIMO compared
and 64T use the same antenna size, whereas 8T uses twice as to remote radios but gains from 64T64R compared to 32T32R
large antenna as 4T, and 32T uses twice as large antenna as 8T. are significantly lower. The main reason that 32T32R performs
almost as well as 64T64R is that the spread of users in the
The largest gain in all three cases is seen by going from 8T vertical domain is much smaller, so larger subarrays can be used
remote radio to 32T Massive MIMO. The reason is that Massive and gains from vertical domain beamforming are smaller. This
MIMO facilitates more advanced beamforming and MU-MIMO. effect is more pronounced in suburban than in urban. In these
deployments, the lower complexity 32T32R product has similar
performance as the 64T64R product but is more cost efficient,
which can be a deciding factor when choosing product.Note
also the related discussion of coverage and subarray size on the
previous page [Ch. 2, p. 30].

Spectrum characteristics affecting

the radio solution
Although many aspects of multi-antenna technologies Many radio aspects depend on the wavelength, which decreases
are independent of the carrier frequency, there are some as the frequency increases. This means that the size of some
characteristics that differ between frequency bands that will radio components is smaller as the frequency increases. For
impact the Massive MIMO radio solution. example, antenna elements and element spacings are often in
the order of a wavelength making the physical size of an antenna
The duplex mode (FDD or TDD) impacts the Massive MIMO element smaller as the frequency increases. This frequency
solution in many aspects. From a technology perspective, TDD dependency affects the characteristics of products, and the way
solutions can leverage on channel reciprocity for channel state they are built. As an example, at 700 MHz (with a corresponding
information (CSI), whereas FDD solutions typically need to resort wavelength of about 0.4 m), it is “impossible” to fit more than
to feedback for CSI, see [1, Ch. 8]. The duplex mode also impacts two dual-polarized antenna element pairs with 0.5 wavelength
the building practice. FDD solutions need to be dimensioned for element separation in less than 0.5 m. At 28 GHz (with a
simultaneous full uplink and downlink utilization, making, for corresponding wavelength of about 0.01 m), the corresponding
example, cooling and DL-UL channel leakage more challenging. number is roughly 100 dual-polarized antenna element pairs
Legacy lower frequency bands are typically FDD while new with 0.5 wavelength element separation in less than 0.5 m.
higher frequency bands are TDD. Hence, Massive MIMO solutions are great at higher frequencies,
while conventional radio solutions typically dominate at low
frequencies.
As discussed previously, coverage becomes more challenging

at higher frequencies. The desire to maximize coverage at
higher frequencies together with building architecture choices
has triggered a completely new standardized beamforming
framework referred to as beam management, see [TP4] and
[TP8].
Yet another very important aspect is legacy considerations. Most

FDD low-bands and FDD mid-bands have existing solutions to
consider and compare with when upgrading to 5G or Massive
MIMO. The upgrade must provide an improvement to the existing
solution to be considered. New frequency bands have usually less
in-band legacy, e.g. fixed links or satellites, to consider.

TDD mid-band Massive MIMO is the

foundation of the 5G experience
TDD mid-band Massive MIMO is the No
bu
foundation of the 5G experience
Coverage
The Massive MIMO enabling site
sweet spot reuse
Capacity
~3.5 – 5 GHz for future proof
with large networks
bandwidth Massive
Attributes MIMO
Globally
User experience value
boost
available
Greenfield TCO savings

deployments
The mid-band TDD bands (~3.5 — 5 GHz) are the main options Mid-band is an essential network component to reach the full
for 5G deployments. These bands are chosen to meet the 5G experience and Massive MIMO is the key that unlocks the full
requirements of 5G experiences. Large bandwidths (often potential of the mid-band spectrum. Massive MIMO in mid-band
>100 MHz) can provide a substantial performance boost, and will provide great value to MNOs and end-users in terms of
the relatively good coverage offered by Massive MIMO radio capacity and user experience improvements in a TCO optimized
solutions makes the performance boost accessible to a major way. This will ensure a future proof network and enables new
part of the users using the existing site grid. Mid-band spectrum services. The most important value from Massive MIMO comes
is also globally available which provides economy of scales from the improved coverage that can ensure a close to full
and ensures good device availability. Furthermore, mid-band population coverage using the existing site grid.
deployments are greenfield, which means no in-band legacy
considerations. By looking at current trends, it is evident that leading markets
are prioritizing mid-band deployments and results show very
TDD mid-band is also a sweet-spot for Massive MIMO solutions. good population coverage and significantly improved network
The frequencies facilitate antenna arrays of reasonable size and performance. It is also clear that 32TR and 64TR Massive MIMO
allow a digital frequency-domain Massive MIMO architecture radio solutions dominate mid-band deployments.
that can reap all benefits of Massive MIMO technologies, see
[TP6] and [TP9] for further details. A major advantage of TDD
spectrum is the use of reciprocity-based beamforming, where the
downlink channel can be inferred from uplink channel estimates.
This can, when coverage is good enough, provide more detailed
and accurate channel state information (CSI) compared to
feedback-based CSI, which results in better performance for
advanced Massive MIMO technologies like MU-MIMO.

FDD mid-band Massive MIMO to

strengthen other frequency bands
FDD mid-band Massive MIMO is Notes pag
Title mod
strengthening other bands one right
Capacity
<2.7 GHz
improvement s
Good Uplink
coverage improvements
Massive
Attributes MIMO
Legacy
Support value
higher
considerations frequencies
Strong device Off-load the

dependency lowest band
Most spectrum below 2.7 GHz is allocated for FDD, and are often Most FDD bands use 2T2R or 4T4R and for low-band FDD it is
divided into low-band FDD, ~600-900 MHz, and mid-band FDD, challenging to go beyond that due to site constraints on size and
~1.4-2.7 GHz. These bands have typically been used for a long- weight. At mid-band FDD, however, Massive MIMO becomes
time, serving GSM, WCDMA and LTE, and are today providing feasible. FDD Massive MIMO is applicable already for LTE and
the backbone coverage for voice, mobile broadband and internet further improved with NR. Deployments of FDD Massive MIMO
of things (IoT). Most networks are dimensioned to provide full have started and are in the thousands in many regions such
coverage based on some low to moderately demanding service as North and South America as well as Southeast Asia. The
on, for example, 1.8 GHz (GSM) or 2.1 GHz (3G). applicable bandwidth is ~20 – 65 MHz per service provider.
The low frequency FDD bands have good radio propagation FDD Massive MIMO will bring value to service providers and
properties and are essential for providing nation-wide coverage, end users. For congested deployments that already use all
in particular for the uplink. FDD bands have been deployed available spectrum (e.g. legacy technologies or mid-band
for a long time, and there are legacy aspects that need to be not yet deployed), it is imperative to fully utilize the available
considered. Support of legacy devices and features must be spectrum, and Massive MIMO becomes an important tool. For
ensured. A major difference compared to TDD is that FDD some new uplink heavy services and for matching the predicted
Massive MIMO (only) relies on feedback-based beamforming Gbps downlink speeds, it is expected that uplink will become a
which has a strong dependency on the 3GPP standard and UE bottleneck. Improving mid-band FDD with FDD Massive MIMO
capabilities. Downlink FDD Massive MIMO has great potential will ensure a better uplink. On the same note, improved FDD mid-
for 5G due to good standard and device support. For LTE, it is bands with Massive MIMO will also be important for off-loading
more challenging as there are many legacy UEs with limited the low-bands and for supporting higher frequencies, essentially
support for Massive MIMO. The potential uplink gains from making higher bands perform even better by moving poor users
Massive MIMO are substantial for all radio access technologies or weak links to the FDD bands.
(RATs) as the uplink is less dependent of UE capabilities.

mmWave Massive MIMO unlocks the

potential of the available bandwidth
No
High frequencies Unprecedented speed

(24+ GHz) and capacity potential
Very large bandwidth Capacity

(~10x more spectrum) booster
Massive
Attributes MIMO
For good Off-loading
value
radio lower-bands
conditions
Radio Enables revenues
challenges from new services
High-band, or mmWave, refers typically to the spectrum range Due to the very large bandwidths, mmWave has the potential
24 -50 GHz. All bands are TDD and offer substantial bandwidth to give unprecedented speeds and capacity. mmWave works
in the order of 400-2000 MHz per service provider (~10x more well in good radio conditions, for example, smaller cells or
than many other bands). The large bandwidths mean mmWave deployments with a high degree of line-of-sight (LoS). mmWave
can offer extremely high peak rates and capacity. serves as a capacity booster in targeted areas, such as stadiums,
cities, or traffic hot spots. mmWave can also be beneficial in
The high frequencies (~10x higher than traditional frequency off-loading valuable lower bands, that is, mmWave can serve
bands) pose some challenges. Radio propagation becomes much users that have good enough radio conditions and lower bands
more challenging and ubiquitous coverage cannot be expected can focus more on users with poor radio conditions. mmWave is
even with very capable mmWave products. The coverage also well suited for some key services. FWA is one example of a
challenge at mmWave has made Massive MIMO mandatory. The service that can work well at mmWave. As long as the radio is in
high frequencies, the large bandwidths, and the need for many line-of-sight, mmWave can deliver Gbps speeds at distances up
antenna elements with beamforming both at the gNB side and to several kilometers. Another example is industrial applications,
the UE side to maximize coverage impact also the architecture. where mmWave might be vital to meet the high requirements on
More analog processing is required and gNBs and UEs typically performance, see [Ch. 5.1] for an extended discussion.
use some level of analog time-domain beamforming [TP6],
which essentially means that the gNB and the UE can only
transmit/receive in one direction at a time. This has also
forced a completely new beam management framework to be
standardized. The radio realization of mmWave products is also
challenged by the high frequencies. This will be further discussed
in [TP6] and [TP8].

Three constraints for network evolution

TCO per
Three constraints for network evolution Notes pa
Ease-of-deployment Cost efficiency Energy performance
5G
Energy
4G
3G
2G
Time
Traditional 5G deployment
Remote Massive 5G + Ericsson’s ‘breaking the

radiounit MIMO energy curve’ approach [5]
In addition to fulfilling the performance requirements outlined in The typical service provider investment cycle is typically around
previous slides, the solutions must also meet some deployment 5-7 years. Therefore, it is reasonable to seek optimum product
related constraints to be attractive to service providers. The most solutions for the corresponding time period. In the middle figure,
significant and generally applicable constraints are ease-of- different colors describe different cost contributions.
deployment, cost efficiency and energy performance, which are
explained below and further elaborated in the following pages. Energy performance: Adding more equipment that operates
on new frequency bands requires additional use of energy.
Ease-of-deployment: Having products that are easy to install The cost for the network energy consumption is significant
and fit well into the environments where they are deployed already today and even further increased energy consumption
is of key importance in many cases. In many deployments, when adding 5G to the network is unwanted. Service providers
requirements include physical size, ease of installation, and look for minimum use of energy for 5G equipment and
visual impact. These may be driven by site building constraints, reduced use of energy on other equipment wherever possible.
esthetical requirements, wind load, and site accessibility. In There is also an ever-increasing requirement on reducing
some cases, they represent determining factors in the product the CO2 footprint which also relates to minimizing energy
selection process. In many cases, it can be a difficult process to consumption. New standalone 5G solutions must consume low
seek permission to rebuild a site regardless if the site is owned levels of energy. New site solutions that also include replacement
by a landlord or an infrastructure company where the site lease of parts of the existing solution must minimize the energy
agreement is regulated in site rental fees. For this case, small and consumption for the whole site solution.
discrete product variants will be easier to deploy and can reduce
the site rental fee. These constraints need to be considered in all deployments and
are therefore expected both for Massive MIMO and conventional
Cost efficiency: All service providers have limited CAPEX and Remote Radios. The constraints may also be deciding factors
OPEX budgets. They need to build mobile networks that are cost concerning the product to choose. This is further elaborated in
efficient over time. It is suggested to measure cost efficiency as [Ch. 3] and [Ch. 4].
TCO per delivered capacity, as cost efficiency over the whole
investment period is of importance. To reach cost efficiency
requires an appropriate choice of product for efficient use over a
longer time period.

Integrated site solutions for

ease-of-deployment
Introduce 5G mid-band Leverage latest Massive MIMO Reduced TCO

with zero footprint portfolio with beam-through Addressing rental cost, wind
(reduced footprint) passive multiband antenna load and visual appearance
An integrated site solution with the latest beam-through Reducing the footprint and modernizing existing FDD bands,
technology is a fundamental building block for achieving ease- as well as avoiding an additional antenna panel when adding
of-deployment. It allows adding mid-band Massive MIMO and mid-band to an existing site, reduce TCO substantially. Site rental
modernizing existing FDD bands using multiband radios with is usually based on a combination of the number of antenna
increased performance and reduced overall energy consumption panels, wind-load area and visual appearance. An integrated
as well as reduced footprint in terms of size, weight and wind site solution using beam-through technology addresses all these
load. This solves the frequency expansion challenge which many aspects and lowers the site rental significantly.
service providers are experiencing on many sites.
The beam-through technology enables an architecture where

an “arbitrary” active antenna can be placed behind the passive
antenna. Without beam-through, a tedious joint design of the
active and passive antennas is required to guarantee that the
different antennas do not interfere each other too much. The
beam-through technology means that the latest Massive MIMO
radios can be back-mounted as-is behind the passive antenna.
The key enabling technology for beam-through is frequency
selective surfaces that are tuned to serve as a reflector for the
passive antenna, while being transparent for the underlying
Massive MIMO radio.

Higher degree of integration leading

to smaller, lighter and more energy
efficient products
Efficient hardware implementation—an enabler of the Massive MIMO solutions
Ericsson
Uplink
Booster
Multi-User
Digital front-end processing MIMO
Beamforming processing 16/8 layers
Layer 1 processing
Footprint
and power
efficiency
Ericsson
Interference
Sensing
A high level of integration is necessary to enable an energy There is a strong industry momentum towards developing
efficient and powerful Massive MIMO solution with a rich feature custom silicon to reap the benefits in terms of performance,
content whereof many require very high processing capabilities. efficiency, size, weight, etc., and there are many reasons for
this development. One reason is that Moore’s law is slowing
Ericsson Silicon, [6] is at the heart of Ericsson Radio System down. Hence, improving performance and energy efficiency
including Massive MIMO, and there is a range of silicon going forward means focusing more on the architecture and
solutions for different applications. It is modular, and the same implementation, than simply rely on raw transistor power.
architecture is reused across the silicon range. Ericsson Silicon
builds on the so-called Ericsson many core architecture (EMCA), The Ericsson Silicon is the enabler for the rapid development
which comprises hundreds of digital signaling processes (DSPs), of the Massive MIMO products and many of the novel Massive
coupled with some state-of-the-art hardware acceleration MIMO features. For example, the type of advanced multi-
units. On top of this there is a tight co-design between hardware dimensional number crunching used in Ericsson Interference
and software. The use of custom silicon focuses on front-end, Sensing requires substantial dedicated real-time parallel digital
beamforming and layer 1-2 processing, but other processing is signal processing (DSP) and hardware accelerator processing
also considered. capacity, only possible with that latest generations of Ericsson
Silicon™. See [TP6], [TP7.3] and [TP9] for further discussions.

Cost efficiency matters

for investment decisions
Cost efficiency matters for investment Ch
No
decisions Th
Q:
Pre
TCO for infrastructure OPEX CAPEX ca
Cost per site TCO per site
Small differences in TCO for remote
radios and Massive MIMO
Remote radio Massive MIMO

Capacity gain
Capacity gain
Capacity needs vs. product potential
Massive MIMO efficiency is scenario
dependent
TCO/GB
TCO/GB
Cost per delivered capacity
Massive MIMO has superior total cost
per capacity
A per-site total cost of ownership (TCO) analysis is an important Second step is to determine the offered capacity of the products.
step in the process of deciding a recommended radio solution. As illustrated on [Ch. 2, p. 31], the capacity benefits of Massive
TCO calculations depend on many local factors specific to the MIMO over a conventional solution depend on many aspects.
service provider. This slide illustrates three steps of the TCO In many deployment scenarios it is reasonable to assume ~2-3
analysis that highlights analysis that highlights that what times more capacity from Massive MIMO than a conventional
matters most for investments decisions is TCO per delivered system. The figure illustrates that 64T Massive MIMO provides
capacity. twice the capacity compared to 8T remote radio in an urban
mid-rise deployment scenario.
The figure illustrates the three main steps in the TCO analysis
where an 8T remote radio product is compared with a 64T Third step is then to normalize the TCO results from the first step
Massive MIMO product in an urban deployment scenario using with the predicted capacity gain from the second step. As the
mid-band 3.5 GHz. TCO for Massive MIMO is only slightly higher than for remote
radio, but Massive MIMO provides roughly twice the capacity
First step is to run a traditional TCO analysis of the cost parts, compared to the remote radio, the TCO per capacity is reduced
where both CAPEX and OPEX are evaluated over a typical with more than 40% if Massive MIMO is used.
depreciation time. Although the upfront CAPEX for Massive
MIMO is higher than for conventional RRU solutions, this step The analysis illustrated here indicates that in many cases
often demonstrates that the TCO difference between deploying Massive MIMO is a cost efficient solution to provide capacity for
mid-band with a conventional RRU and a more advanced mid-band deployments. Note that a TCO analysis also may be
Massive MIMO solution is rather small. This is mainly because done to motivate a Massive MIMO deployments for coverage
the conventional solution has higher cost related to baseband purposes. More details on the TCO analysis can be found in
capacity, network roll out, site rental and other parts. [TP9.4].

Our energy performance journey No chang
Early (LTE) Today (NR) Future
Power Focus on thermal

Power Power Size and weight
Micro-sleep More band-

Lean NR width. More
standard RF power
Analog
Analog (PA) More MIMO
Higher data Analog
Digital
Digital (BB) rate Digital
Load Load Load
Focus on power Focus on energy Operational cost

amplifier idle mode proportional computing and carbon emission
In existing mobile systems, where energy consumption has low In order to improve the load dependency, micro-sleep
dependence on the load and traffic growth is high, an energy- transmission is introduced, an energy saving feature that
efficiency metric is not particularly useful. Even if no effort is deactivates and reactivates power amplifiers (PAs) in
made to reduce energy consumption, traffic growth will cause microseconds. Micro-sleep transmission is effective whenever
energy efficiency to improve from one year to the next. A more there is no transmission from the base station. However, it could
useful approach is to focus on the energy performance of the be even more effective if the LTE standard would not require
system, where energy performance is defined as the minimum mandatory and always-on reference signals that cannot be
achievable energy consumption for a set of performance turned off, even when there is no traffic.
requirements (user throughput, capacity, latency and so on).
The term energy performance broadens the focus beyond The physical layer of NR was developed to ensure an ultra-lean
energy per bit to total energy consumption, highlighting the design [7]. This design makes features such as micro-sleep
similarities between achieving high system performance transmission much more efficient, and the effect on live networks
and low energy consumption. Reducing energy consumption is already significant [8]. Micro-sleep transmission in NR
without also considering performance is pointless as in that dramatically decreases the energy consumption in the analog
case all equipment can just be turned off to have zero energy radio parts of base stations. While NR has superior support
consumption. for micro-sleep transmission it also features wider system
bandwidth and more antennas, which increase the energy
The network energy performance journey of mobile networks: consumption in digital components. The energy consumption in
The graph on the left, shows the power consumption of a typical a NR base station is no longer dominated by the analog power
LTE base station in the period 2010-2020. The graph in the amplifiers (as in LTE) and the energy consumption of the digital
middle shows the power consumption of a typical New Radio components can no longer be ignored. Going forward, improved
(NR) base station today. The graph on the right shows the scalability in digital components will become increasingly
projected power consumption of a mature NR base station and important.
6G base station.

Summary No
● New services, improved user experience and increased

network traffic load drive network requirements
● 5G networks will require significant enhancements

of coverage, capacity and user throughput
● 5G networks must be easy to deploy and be

cost- and energy efficient
● Each site is unique with respect to performance

requirements and deployment constraints
● Characteristics of different frequency bands, such

as FDD, TDD mid-band and mmWave, will impact
the radio solution
A mobile network that delivers 5G experiences must not only lthough many aspects of Massive MIMO are independent of
A
enhance user experiences but also handle exponentially the carrier frequency, there are some characteristics that differ
increasing traffic loads. This is achieved by increasing capacity, between frequency bands that will impact the Massive MIMO
enhancing user throughput, and improving coverage on existing solution.
as well as on the new and higher 5G frequency bands (to make
use of the full capacity on the 5G band in the whole coverage The requirements and constraints outlined above serve as input
area using the existing site grid). to the design of radio and feature solutions portfolios. As can
be seen, the capabilities of Massive MIMO outlined in Chapter
In addition to these network performance requirements, 1 meet many of the requirements outlined here in Chapter 2.
there are also some constraints to consider when choosing Hence, Massive MIMO is an attractive solution for many 5G
the appropriate product solution for a specific site. Some deployment scenarios. The principles for how a portfolio can
environments put requirements on ease-of-deployment, e.g. be structured and examples of how it is done by Ericsson are
requirements on small size, low weight and/or low visual impact. outlined in the next Chapter 3.
Cost efficiency and energy efficiency are general constraints
that all service providers need to consider. All service providers
have constraints on both CAPEX and OPEX and for most service
providers there are both regulatory and legal requirements,
political goals or incentives and corporate strategies to
limit energy consumption and the production of CO2. These
constraints are thus highly relevant evaluation criteria in the
process of product selection.

3. Massive MIMO
solutions
3. Massive MIMO solutions

Ericsson | The Massive MIMO handbook 3. Massive MIMO solutions 43
Outline
• Radio architectures differences Hardware features

between Massive MIMO radio
and conventional radio
• Ericsson Massive MIMO radio
portfolio
Software features
Massive MIMO
• Ericsson Massive MIMO feature mmWave
portfolio
Remote radio
unit (RRU)
RAN Compute Passive Interleaved Massive MIMO Massive MIMO

—baseband unit antenna / Hybrid FDD TDD mid-band
Ericsson Radio System (ERS) is designed to meet the his chapter is outlined as follows. The first section describes
T
requirements of past, current and future mobile networks. differences in architecture between Massive MIMO and
It provides a versatile toolbox that includes Radio, RAN Compute conventional RRUs. Thereafter follows two sections describing
[9], Radio System Software/Solutions, Transport, Antenna the Massive MIMO radio solution and Massive MIMO feature
Systems and Radio Site System. Here, three key components solution, respectively. Discussions on how the radio and
of a radio solution, viz. baseband*, radio units, and antennas, feature portfolios can be segmented to address the network
are in focus. requirements and constraints are also provided
* In the Ericsson implementation, the baseband functionality is

implemented either as a baseband unit, shown in the figure, or
as a cloud-based solution.

3.1 Massive MIMO solutions

—solution architectures

Different functional distribution:

Remote radio units vs. Massive MIMO
radio units
Mid-band and mmWave
Analog front-end processing
Massive MIMO radio unit
Digital front-end processing
Typically, Analog front-end Beamforming processing
2T2R – 8T8R RF cables Layer 1 processing
Digital front-end Layer 2 processing
Remote radio unit
Analog front-end Beamforming
Digital front-end Layer 1
CPRI or eCPRI eCPRI

Massive MIMO features
distributed between radio
Baseband unit Baseband unit and baseband units.
Layer 1 Beamforming Layer 1
Layer 2 Layer 2
Massive MIMO includes hardware and software, as described is a reason for also the remote radios to use eCPRI. The data
on page [Ch. 1, p. 13]. Functional distribution in Massive MIMO compression ratio that eCPRI provides is on a high level related
solutions is somewhat different than in conventional solutions. to the fact that data rate over eCPRI typically scales with the
The reason for this is that a Massive MIMO radio unit has more user traffic bitrates, in contrast to CPRI, where the data rate is
radio chains and therefore a need for transferring substantially fixed and scales with the number of antennas. The reduction of
more information to the baseband unit. For example, a 64T64R front haul bitrates can be to the order of ten, or even higher for
mid-band Massive MIMO radio unit would require 16x the mmWave products that have a large number of antennas.
capacity of the common public radio interface (CPRI) link than a
4T4R radio unit. However, even though the functional distribution In addition, to avoid extremely high bit-rate links, there are also
is different, the same Ericsson baseband solution can handle benefits of reduced total processing time, such as processing
remote radios and Massive MIMO radios. in radio, transmission time from radio to baseband, processing
in baseband, transmission time back to the radio, processing in
There are two ways of solving the interface issue: either provide radio, by locating functionality in the radio. Certain time-critical
more CPRI capacity to the baseband or move functionality and transmission-volume-consuming data handling is in the
to the radio to aggregate information already in the radio. radio’s uplink receiver (also called the Uplink Booster, [10]), which
The solution chosen by Ericsson is to move some parts of the greatly benefits from having access to instantaneous channel
Massive MIMO functionality to the radio unit, which is made information.
possible by using the eCPRI interface (CPRI with support for 5G
fronthaul). eCPRI is an evolution of CPRI that uses the packet- The overall functional distribution is the same for mid-band and
based Ethernet protocol and enables more flexibility and data mmWave Massive MIMO. However, there are differences both in
compression, such as moving the beamforming function to the the hardware solutions and the software features, which will be
radio unit. The use of Ethernet enables switched fronthaul which described in later sections.

Design efficiency enabled by Ericsson Silicon
Baseband unit Massive MIMO radio unit Superior

Layer 2 Layer 1 eCPRI Digital Analog front- architecture
front-end end
BB ASIC (with DSP), CPU Radio ASIC Analog HW Ericsson Silicon
Layer 2 Layer 1 Beamforming

Beamforming Full IRC receiver
BB ASIC (with DSP), CPU Layer 1 DL beamforming
BB ASIC (with DSP), CPU

CPRI or
eCPRI
eCPRI
Analog front-end processing Remote radio unit
Digital front-end processing
Digital Analog
Beamforming processing
front-end front-end
Layer 1 processing
Layer 2 processing Radio ASIC Analog HW Baseband unit
Hardware platform
The digital functions of the baseband units and the radio units he trend with more radio chains and more total configured
T
are implemented by software running on application-specific bandwidth (TCBW) is driving the processing requirements.
integrated circuits (ASIC). Compared to conventional radio To cope with this increase in complexity and to keep the power
units, TDD mid-band Massive MIMO radio units have the same consumption on a manageable level, Ericsson has developed
type of ASICs as the baseband units, including the software a set of custom ASICs, Ericsson Silicon. Inhouse development
that implements a subset of the baseband functions that are makes it possible to target the design of the silicon for best-in-
implemented in the baseband unit for remote radios. class algorithms.
In addition to the baseband functions, the Massive MIMO radio From a portfolio perspective, the radio portfolio comprises the
also includes ASICs to perform digital radio processing. Since the hardware of the Massive MIMO radios (AIR products), and the
Massive MIMO radio has many more radio chains than a remote software features, Massive MIMO features and other common
radio* there is considerably more processing needed. Thus, there features, comprise the software, which is located both in the
is more hardware and more functionality in the Massive MIMO Baseband and the Radio. The remainder of the chapter will
radio compared with a remote radio. The Massive MIMO radio is elaborate on how the radio portfolio and the Massive MIMO
a more complex but also a more capable product. features can be designed to cover all requirements in the 5G
network.
*) Remote radio units using the eCPRI interface typically include

a small part of the Layer 1 processing (not shown in the picture),
e.g., the FFT and cyclic prefix processing.



—Massive MIMO radio

Site classification — Traffic load and

deployment environments No
Traffic load
dense urban
suburban
rural
t
en
m
In on se
c rea vir re
a
sin t en nc
g tra en yi
ffi y m xit
cl
oa plo ple
dp De co
m
d
er
sit b uil
e
e re
Sites Sit
Service providers need to evolve their networks to meet the Another key factor determining the effectiveness of different
future performance needs in a cost- and energy efficient way. Massive MIMO configurations is the deployment environment,
In this process, it is vital to understand the network reality and typically classified as rural, suburban, urban or dense urban.
how it affects the choice of hardware and software solutions. The site density (inter-site distance) is important deployment
Each site is unique in terms of traffic load, expected traffic parameter that affects coverage. Also, the distribution of users
growth, deployment environment, ease-of-deployment, etc. in the vertical domain is important as shown in [Ch. 2, p. 30-31].
To guide the design of a radio product portfolio and to The deployment environment corresponds to one of the axis in
understand where to use each product, the requirements on a the figure above.
site level are now mapped to different deployment scenarios
and traffic load levels. The load categories are then mapped onto the different
deployment scenarios, ranging from rural to dense urban high-
Traffic load and expected traffic growth are important factors rise. Note that sites with different load characteristics are found
when deciding how to evolve the network. Many of today’s in all types of deployment environments. A third factor, which
networks have a traffic load distribution over sites that look is usually closely related to the deployment environment is site
according to the figure above, where red, blue and green rebuild complexity which in turn relate to ease-of-deployment
correspond to sites with very high, moderate, and low traffic load, as described on [Ch. 2, p. 36].
respectively. Hence, most sites experience low traffic load and
rather few sites have very high traffic load. The network realities are central in the network evolution
process and also key for deciding how to develop the hardware
and software product portfolios.

Ericsson radio product segmentation
Ericsson radio product segmentation
Capacity Compact Coverage
Targeting urban (incl high rise) Targeting urban and suburban Targeting urban to rural deployments
to suburban deployments deployments with large inter-site distance
Mast and rooftop deployment Rooftop and wall mounted deployment Mast and Rooftop deployment
Capacity (bandwidth, multi-band) Footprint optimized capacity, EIRP and bandwidth (multi-operator RAN)
and EIRP optimized addressing cost, ease-of-deployment optimized coverage, addressing TCO
and energy consumption and sustainability
The radio portfolio of Ericsson has been categorized into three The second category, Compact, targets deployments where
product segments, capacity, compact and coverage, where ease-of-deployment is important. It prioritizes TCO and
the products in the respective categories target different ensures that mechanical properties are in line with common
purposes and deployment scenarios, and thereby have different site constraints, such as size and weight. Still, the products
characteristics. The most important design parameters used for can provide substantially performance gains compared to
defining these segments are: radio parameters such as number conventional remote radios in many deployment scenarios.
of radio chains, output power, number of frequency bands and
total configured bandwidth (TCBW), antenna parameters The third category, Coverage, targets deployments with larger
such as antenna size, number of antenna elements, subarray inter-site distance like suburban or rural. It has high EIRP (high
structure, cost and building practice parameters such as size and output power and many antenna elements) to ensure coverage
weight. and typically fewer radio chains to lower total cost of ownership
(TCO) compared to the Capacity segment. Fewer radio chains
The first category, Capacity, targets the most capacity is well motivated as targeted deployments have less spread of
demanding sites and provides superior performance in all users in the vertical domain, hence less use of vertical domain
deployments ranging from dense urban high-rise to rural. It beamforming. Note also that large antennas are desired to
supports multiple frequency bands, large bandwidth, high maximize both downlink and uplink coverage.
effective isotropic radiated power (EIRP), and many radio
chains facilitating superior horizontal and vertical domain It needs to be highlighted that capacity, compact and coverage
beamforming. refer here to product segments, and should not be mixed up
with network properties such as coverage, capacity and user
throughput. That is, there is not a one-to-one mapping between
the segment names and ‘corresponding’ network properties.
For example, products from the capacity segment will provide
superior overall performance including coverage, capacity
and user throughput.Also to be noted is that TDD mid-band,
FDD and mmWave have different characteristics and different
applicability and hence, the product portfolios are different.
See also Start simple with the Ericsson Radio System, [11] for
a general introduction of the Ericsson radio portfolio.

Portfolio mapping in network deployment

Lo
Traffic load
per site
Capacity
Optimized for capacity
Compact
Optimized for site deployment
Coverage
Capacity
Massive
Compact
Optimized for larger inter-site distances
MIMO
Remote Coverage
radio
Sites
The aim of the portfolio is that the product segments shall meet There is no clear-cut classification, so the mapping acts as a
the performance requirements on site level in different parts high-level guide. Products targeting highly loaded sites need
of the network given relevant constraints. The radio product to provide superior capacity meaning many frequency bands,
segments introduced in the previous slide can be roughly large bandwidths and often excellent multi-antenna support. In
mapped to site classes introduced above in [Ch. 3.2, p. 49] these deployments, Massive MIMO typically outperforms remote
according to the figure. radio. Products targeting site with low load can use slightly less
capable products but instead excel in terms of TCO and ease-of-
In the deployment environments dimension, deployment. In this case, remote radio might be a costefficient
the Coverage segment is primarily intended to address alternative to Massive MIMO.
sites in rural and sparse suburban environments and
the Capacity segment addresses sites in the suburban, urban Similarly, products in dense urban high rise deployments benefit
and dense urban environments. The Compact segment is a from highly capable beamforming in both horizontal and vertical
complement to the portfolio for sites where there are restrictions domain, many Ts and Rs, whereas products in rural deployments
on deployment, e.g. size, weight, wind load, visual impact, and/or need less vertical domain beamforming, fewer Ts and Rs.
cost efficiency. Products in rural deployments also need to have large EIRP
(large antennas and higher power spectral density) to achieve
In the traffic load dimension, Massive MIMO addresses primarily good coverage.
sites where the traffic is medium to high and remote radio
addresses sites where the traffic is low and is expected to remain One thing to note is that sites with all different load levels can
low throughout the lifetime of the radio. occur in all deployment scenarios, e.g. highly loaded sites in rural
deployments. A highly loaded site in rural requires a slightly
different Massive MIMO product compared to dense urban
high-rise, e.g. lower degree of vertical domain beamforming in
rural compared to dense urban high-rise.
Depending on market requirements, the product availability

is different for different bands and hence all products are not
available for all frequency bands. The portfolio will however be
more complete over time as market requirements develop.

TDD mid-band
Massive MIMO portfolio
AIR 3219
Evolution
Coverage
EIRP Higher EIRP
optimized Dual band General
320 W
200 MHz Higher bandwidth
AIR 3227 Interleaved
Compact AIR 3268 AIR 3255 Hybrid AIR Interleaved AIR
Size/weight/
active and
Footprint
passive energy reduction
optimized
capacity dual band
200 W 200 W 200 W PA efficiency and
200 MHz 200 MHz 200 MHz power pooling
Capacity AIR 3229 AIR 3258 AIR 6419 AIR 6428 AIR 6476 Multi band
Fronthaul evolution
Bandwidth Higher EIRP
and EIRP
optimized New spectrum
AIR 3218
400 W 320 W 320 W 400 W 400 W AIR 6418
Multi band 400 MHz 200 MHz 400 MHz 600 MHz
There is a fast evolution of mid-band Massive MIMO products, 19 kg (for B42 band variant). In the Compact segment, products
and the Ericsson mid-band Massive MIMO portfolio has already have been optimized with respect to its mechanical properties
gone through multiple hardware generations. Each generation and TCO for the operators with constrained site deployment in
contains improvements in many areas, e.g. bandwidth and focus. As an example, AIR 3255 has 200 W output power, very
effective isotropic radiated power (EIRP) have increased small footprint and the weight is only 12 kg. The product is a
substantially, while the weight has reduced dramatically. As an good compromise between high capacity and small size, which
example, one of the latest capacity Massive MIMO radio, AIR is useful in many network deployments around the world within
6428, has more than six times the bandwidth, more than four urban and sub-urban environments. The cost effectiveness, TCO
times the EIRP and still less than half the weight compared to the per capacity, is very attractive for many service providers. In the
first generation Massive MIMO product. Today there is a variety Coverage segment the main focus is to provide high coverage
of Massive MIMO products that are optimized for different and lower TCO in wider areas with limited capacity needs. The
segments and deployment scenarios. One important component high requirements on coverage require high EIRP (320 W total
for making this evolution possible is the use of optimized output power, 192 Antenna Element). Due to the lower demand
application-specific integrated circuits – the Ericsson Silicon - for for capacity and lower demand for beamforming in the vertical
beamforming and physical layer processing (see [5]). It should dimension a 32T32R rather than 64T64R is sufficient. Hybrid
also be highlighted that these improvements do not only make and interleaved products are discussed on [Ch. 3.2, p. 57].
the hardware specifications much more attractive, but they also
enable more novel software algorithms. To the right in the picture, examples of how the product segments
will evolve over time are provided. Some improvements are
A summary of the Ericsson Massive MIMO portfolio, as of segment specific. For example, the compact segment focuses
2023Q1, is illustrated in the slide. One Massive MIMO product on evolving the interleaved active and passive products and
in the Capacity segment is AIR 6419 which provides high adding dual-band support. Other improvements are common to
EIRP (320 W, 192 Antenna Element) and high beam steering all segments, e.g. mechanical reduction and energy consumption
capability in the vertical domain (64T64R). The weight is below improvements.

Power consumption scalability of a TDD

mid-band Massive MIMO radio Min
Pow
Idle
AIR 6419
Maximum power consumption 100%
Average power consumption 45%
Cell idle 35%
Cell blocked 25%
Deep sleep 10%
The figure illustrates the power consumption scalability of The first step is to reduce the cell idle power consumption.
AIR 6419 compared to the maximum power consumption. This is achieved by turning off PAs on a symbol level when
no traffic or signals need to be transmitted. Smart scheduling
The maximum power consumption of the radio serves as an decisions can be made to maximize the idle time with little
important input for site dimensioning but is of less interest for impact on the network performance.
a TCO driven analysis. A better value for a TCO analysis is the
average power consumption that reflects 24/7 operation with The second step is to take down or block the cell. In this step,
energy efficiency features enabled. all PAs are completely shut down.
As discussed on [Ch. 2, p. 40], the software and hardware The third step is to enter deep sleep, where also digital
operation at low traffic load is very important from an energy components are shut down. This results in larger energy savings
performance perspective as many cells have very low utilization but also longer activation times.
most of the time, and this is when the lean design of NR can
have a huge impact. There are large traffic load variations over time and over sites,
and Ericsson’s hardware and software leverage on these
variations to provide high-performing networks in an energy
efficient way. Power should only be consumed when needed.

FDD mid-band Massive MIMO portfolio
AIR 3246 AIR 1641 AIR 3283 AIR 3284 Evolution
Multi band
Wide band
Weight
Capacity reductions
Optimized
antenna
size
Single band, 160 W Dual band, 2x160 W Dual-band, 320 W Triple-band, 520 W
32T32R 16T16R shared 32T32R shared 32T32R
CPRI interface and solution architecture similar eCPRI interface and Massive MIMO
to remote radio [Ch. 3.1] solution architecture [Ch. 3.1]
There are some characteristics that impact the FDD Massive Massive MIMO. Due to the importance of time to market, the
MIMO products and tend to make them larger and heavier than solution architecture is similar to remote radios with a CPRI
corresponding TDD products. Many radio components become interface and all beamforming processing in the baseband unit,
larger at lower frequencies, and FDD operation implies full see [Ch. 3.1]. AIR 3246 is a single band unit with 160 W total
simultaneous uplink and downlink utilization. These aspects output power and a four column antenna, providing both vertical
have some implications. For example, RF filters become large and horizontal beamforming capability. AIR 1641 is a dual
and bulky to satisfy out-of-band requirements on unwanted band unit with an eight column antenna providing horizontal
emission or interference towards/from other bands or the beamforming capability and remote electrical tilt in the vertical
own receiver/transmitter, and Massive MIMO solutions have domain. The wider antenna provides narrower beamwidths,
many transceiver chains, hence many RF filters. Passive enabling better sector splitting than AIR 3246. AIR 3283 is
intermodulation (PIM) is one problem that becomes more a second-generation dual band 32T32R product. It uses the
challenging when more and more frequency bands are being same solution architecture as the TDD mid-band Massive MIMO
deployed. Another example is cooling, that also becomes more products, see [Ch. 3.1], which provides a number of advantages
challenging when product dimensioning needs to cater for full over the previous generation products and facilitates a more
simultaneous uplink and downlink utilization. Cooling is a major efficient large-scale network rollout. The weight is reduced by
contributor to size and weight. Yet another aspect impacting ~50% compared to previous generation. Second generation
the properties of FDD Massive MIMO relative to mid-band TDD product improvements are enabled by Ericsson Silicon and the
Massive MIMO is that FDD Massive MIMO products are less eCPRI interface. By moving beamforming processing to the radio
mature and have focused on time to market rather than cost unit, the required baseband capacity is reduced, which improves
optimization. Many aspects improve for each generation. the TCO. Furthermore, a custom design wide band RF power
amplifier enables the total output power to be shared across two
All FDD Massive MIMO products belong to the Capacity frequency bands, which makes it possible to adapt to asymmetric
segment. The main purpose of FDD Massive MIMO is to provide spectrum allocations. AIR 3284 is a triple-band product with
additional capacity over traditional remote radio products. 140 MHz total carrier bandwidth, 520 W total output power and
All FDD Massive MIMO products support LTE, NR and mixed a weight of only 47 kg.
mode operation with LTE and NR in the same band. The first
generation products were introduced to demonstrate FDD

mmWave portfolio
AIR 5322 AIR 5343

Capacity eMBB
High EIRP FWA
Hot spot City macro sites Street sites Macro tower
• Performance • Leverage • Densify for best • Coverage

and capacity existing sites coverage targeting FWA
customers
• Indoor and • Spotty • Urban and
coverage suburban
Compact outdoor
AIR 1281 Streetmacro
Low weight and size
6705
with or without
integrated baseband
The high frequencies (~10x higher than traditional frequency operating bandwidth of 800 MHz, an EIRP of ~53-56 dBm and a
bands) pose some challenges for the mmWave products. Radio form factor of 7 liter and ~8 kg. Streetmacro 6705 belongs to the
propagation becomes worse, and products need to maximize Compact segment and is a complete base station with integrated
coverage by supporting large antennas, hence many radio high-capacity RAN Compute (baseband). This enables a smooth,
chains. Many radio components shrink also as the frequency fast and cost efficient rollouts of mmWave. It has an operating
increases, and small form factors can make, for example, bandwidth of 800 MHz, an EIRP of up to 60 dBm and a form
cooling challenging. The power efficiency and the receiver factor of 10 liter and 13 kg.
sensitivity decrease with frequency, which also affect the
thermal dimensioning and the uplink coverage. Phase noise As outlined in Chapter 2, the preferred mmWave deployments
is also becoming more problematic as the frequency increase. are small cells with high traffic load and pre-dominantly
The large bandwidth and the need for large antennas, hence line-of-sight propagation conditions. The current trend is that
many radio chains, impact also the radio architecture. More service providers deploy mmWave for primarily eMBB and
analog processing is required to reduce cost and complexity, and FWA services in areas with high traffic load. All mmWave
mmWave products use analog time-domain beamforming. See service providers deploy in hot spot areas as stadiums, airports
also [TP6] and [TP8]. or train stations. In these areas, mmWave delivers high user
peak rates and cost efficient capacity. Many service providers
There are two mmWave product categories, roof-top macro, also deploy mmWave radios on existing city macro sites, typically on
referred to as antenna integrated radio (AIR), and Streetmacro. rooftops. The addition of a small mmWave radios add significant
AIR 5322 belongs to the Capacity segment and has an operating capacity to the site. mmWave coverage using existing sites will
bandwidth of 800 MHz and a total Effective Isotropic Radiated often be spotty, but with a good multi-band network, users will
Power (EIRP) of 59-62 dBm depending on configuration and have a consistent experience. mmWave on dense street sites is a
frequency band. The form factor is ~7.5 liter and ~8.0 kg. The rather common deployment in the US that can result in full street
form factor is ~7.5 liter and ~8.0 kg. AIR 5343 also belongs to coverage. FWA also fits well with mmWave. By using mmWave
the Capacity segment, a total carrier bandwidth of 1600 MHz together with Massive MIMO on mid band, a FWA service with
and a total EIRP of 70 dBm and a form factor of 19 liters and fiber like performance can be delivered very efficiently using the
16 kg. AIR 1281 belongs to the Compact segment and has an existing site grid.

Comprehensive radio portfolio
TDD M-MIMO mid-band

TDD M-MIMO mmWave
FDD M-MIMO
The picture above illustrates the entire radio portfolio (as of When considering other deployments than traditional macro
2023Q1) and specifically the Massive MIMO products available networks, there are also the micro and street segment containing
for TDD mid-band spectrum range, marked in yellow, the FDD radios and AIR products (built in antenna) to be used ‘below
products, marked in red, and the mmWave products, marked rooftop deployment’ and in special hotspot deployments. For
in purple. For further descriptions on how to build radios and dedicated indoor deployments, there are variants of Ericsson
specifically Massive MIMO, see [12] and [13]. Radio Dot System (RDS), where single or multi-band RDS can be
used for efficient mid-band indoor deployment.
When service providers are considering to deploy mid-band
spectrum, new radios are needed. Massive MIMO products When finding a suitable mid-band solution, the entire radio
provide higher capacity than remote radios. They also provide product portfolio should be considered. The scenarios where
better coverage which is required to get mid-band coverage Massive MIMO products will be the best long-term solution will
using the existing site grid, [Ch. 2, p. 27]*. For site reuse in be discussed in Chapter 4.
macro network deployment with limited capacity demand
and modest capacity growth, conventional remote radio, with * The complete Ericsson Radio System portfolio has a lot of
2TRX up to 8 TRX that are connected to passive antennas, is Massive MIMO product variants. The variants come from
a good complement to Massive MIMO. Similar as for Massive market requests that define band variants and properties
MIMO, remote radios are categorized into the product segments per deployment scenarios and are grouped into our product
capacity, coverage and compact. The remote radio with 8T8R is segments. The result of market-driven product requirements
found in the Capacity segment and remote radio 2T2R and 4T4R is band combinations that do not have all different product
in the Compact segments. These radios need to be connected to segments realized. Therefore, it is important that service
passive antenna in order to get a complete radio site solution. providers engage with Ericsson and share their plans and
requests to secure that Ericsson will plan for new product variant
in time for the demand.

Add mid-band with Massive MIMO

and remain with only one antenna after
Add mid-band with Massive MIMO and remain
expansion
Ch
No
with only one antenna after expansion
Hybrid AIR Massive MIMO
● Combines low band antennas with a Massive MIMO
on top, utilizing the wide range of passive antenna modules
from Ericsson Antenna Systems (EAS) portfolio
FDD band combinations
● Option to use variants of 32T32R and 64T64R radio modules
from the Ericsson portfolio
Interleaved AIR Massive MIMO

● A single antenna form factor that uses interleaving beam-through
technology for maintained low band antenna properties for same FDD band combinations
height of antenna utilizing full length
● Enable future expansion to Massive MIMO mid-band within
the same site lease contract
The products presented on [Ch. 3.2, p. 52] assume that a new The main advantages of the combination of an active/passive
product is introduced to cover the requirements on mid-band radio solution as described here is of course ease-of-deployment
without any concern of already existing equipment on site. and TCO. For example, this allows service providers to simplify
the renegotiations on mid-band expansion on the site where the
It would be highly attractive in many deployments to be able to final solution will only contain one antenna radome supporting
have one product that adds Massive MIMO on mid-band, and both FDD modernization, additions of new bands and addition
at the same time modernizes existing FDD bands. Hence, one of mid-band TDD with a Massive MIMO solution.The product of
product that replaces existing products and adds Massive MIMO choice for service providers depends on cost and performance,
mid-band support. Ericsson has two product variants to support but both product variants are key solutions for service providers
the vast demand of different product characteristics on FDD that need expansion on site where only one antenna per sector
antennas on global scale. In the product called Hybrid AIR, there is allowed, or for other reasons, where only one antenna is
is a modular building practice that maximize the selection of preferred.
Massive MIMO AIR products in combination with large number
of passive antenna models for FDD. A more integrated solution is
also designed where the full length of the total antenna product
can be used both for mid-band TDD and the low band FDD parts.
This product is called Interleaved AIR where the top part of the
FDD antenna is transparent towards the TDD Massive MIMO
radio embedded behind the low band part of FDD antennas.


—Massive MIMO
features

What is a Massive MIMO software feature?

No
RAN Software feature portfolio
Massive MIMO
specific features
FDD mmWave RAN features

specific specific
Mid-band
specific
Conventional radio
specific features
The RAN software portfolio consists of a wide range of features. Most Massive MIMO features are applicable to FDD, TDD mid-
These features are typically agnostic to the radio being used. This band and mmWave. There are, however, some Massive MIMO
means that they can be used on conventional as well as Massive features that are specific to FDD, TDD mid-band and mmWave.
MIMO radios. However, Massive MIMO adds the capability to The reasons for this feature split are primarily related to standard
utilize a new set of software features that are Massive MIMO aspects, device issues or architecture and implementation
specific (or features which substantially increase in benefit when concerns.
using a Massive MIMO radio). These features are referred to
as Massive MIMO (specific) features, and they will be further This chapter starts by a high-level description of features and
elaborated on in this chapter. aspects that, to a large extent, are common for FDD, TDD
mid-band and mmWave. The description is written from a TDD
For Ericsson products, most software features can be used with mid-band perspective. In the end of the chapter follows a short
most radios presented in [Ch. 3.2, p. 56]. Some features, like description of some key differences and specific features for FDD
some Massive MIMO-specific features, may require hardware and mmWave.
support and can only be used with Massive MIMO products.
The Massive MIMO features are described in further detail in
[TP7] and the performance that typically can be achieved with
the features, respectively, is illustrated for a number of cases in
[TP9].

What is a Massive MIMO software feature?

No chan
Different permutations
Network needs to Multi-antenna Enablers of the multi-

address? technique? antenna technique?
• Improve coverage • Beamforming • Channel knowledge Unique
• Increase throughput
• Nullforming • UE 3GPP capability Massive MIMO
• Increase capacity
feature
• Reduce interference • Spatial multiplexing • Massive MIMO radio
• Improve energy capability
efficency
Considerations
• Necessity to • Performance • Cost of aquisition

solve problem
• Tradeoffs • Availability
• Tradeoffs • Alternative solution
A Massive MIMO feature can be described by a combination of ll Massive MIMO features can essentially be described by
A
three different categories: applying a combination of the three basic multi-antenna
techniques, beamforming, nullforming and spatial multiplexing,
1. Which network requirement does the Massive MIMO feature to some physical channel or signal, using some available channel
intend to address? For a discussion on requirements, see knowledge in order to solve a certain problem. This may sound
[Ch. 2]. simple, but there are several aspects to consider which results
2. What channel knowledge is available? in a wide variety of potential features. First, the network needs
to acquire channel knowledge in order to know how to perform
3. Which multi-antenna technique, [Ch. 1, p. 15] (or combination beamforming, nullforming or spatial multiplexing. This can be
of techniques) can be applied using the channel knowledge achieved in several ways, but it is important to understand that
(2) to solve the problem in (1). there is always a cost associated with acquiring channel state
information (CSI), for example, increased overhead.
Different permutations of these three categories will yield unique There is also a problem of availability of CSI. Different sounding
Massive MIMO features—potentially with different tradeoffs and and feedback methods are available in the 3GPP standard
applicability to different conditions. and different UEs may have different capabilities and support
different modes. Hence, the network must support several
Firstly, the problems the feature should solve must be Massive MIMO features in parallel. Even if a UE supports a
determined, e.g. improve coverage, increase capacity or increase certain CSI mode, that CSI may not be available at a certain
throughput. Potentially, one feature can solve multiple problems instance in time. For instance, when a UE first connects to a cell,
at once, while in other cases, there may be trade-offs to be made. no channel information is generally available and measurement
For instance, in order to improve energy efficiency, the capacity or sounding configurations will need to be set up via RRC
may be negatively affected. Thus, it is important to assess what configurations implying that there is a lead time before such
network requirements are important for a certain cell at a certain CSI is available to the network. Different sets of Massive MIMO
time. For instance, during off-peak hours, the capacity demand features are needed when no or limited CSI is available compared
in the cell is low and therefore it may be acceptable or even to when CSI is available.
desirable to apply a feature that sacrifices capacity to improve
energy efficiency.

Categories of Massive MIMO features

N
Advanced MIMO for

Coverage extreme performance
• Codebook-based beamforming • 16/8 layer DL/UL MU-MIMO

• Fully flexible UL beamforming Cross-layer optimization • Inter-cell interference management
• Control channel beamforming with other RAN features • Coordinated multi-point (CoMP)
• Spectrum interworking Mobility, RRM, CA, traffic • Automated cell shaping
management, time-critical
services, Quality-of-Service,
self-organizing networks,...
Capacity and Ease-of-deployment,
performance cost and energy
efficiency
• Reciprocity-based DL SU-MIMO • EMF/EIRP regulatory compliance
• Reciprocity-based DL MU-MIMO • Co-existence enablers
• Optimized adaptive beamforming • Sleep mode
• UL MU-MIMO with full IRC • Enablers for efficient hardware solutions
The Massive MIMO features can be put in four different eatures in the category ease-of-deployment, cost efficiency,
F
categories, illustrated with examples in the figure above. These and energy efficiency address aspects not related to network
feature categories will be further elaborated in the coming slides. performance [Ch. 2, p. 36] such as fulfilling regulatory
requirements on electromagnetic field (EMF) emissions, which
• Coverage may require dynamically adapting the power level depending
on the beamforming characteristics of the cell or reducing the
• Capacity and user performance
amount of beamforming that is performed in order to conserve
• Advanced MIMO feature for extreme user performance energy.
• Ease-of-deployment, cost efficiency, and energy efficiency he Massive MIMO segments FDD, TDD mid-band and
T
mmWave will affect which features in the different segments
Coverage features improve either the access coverage or the
that are applicable and sometimes also how the features are
data channel coverage, typically by applying beamforming to
implemented.
boost the signal strength. This can be done both in uplink and
downlink. Capacity and performance features, on the other
hand, typically aim to improve peak user throughput and/or
capacity of the data channels and generally use a combination
of beamforming, nullforming and spatial multiplexing. Advanced
MIMO features target more specific deployments or scenarios
which may have an extreme capacity demand, very high load
or some other special characteristics. For these cases, more
advanced features such as inter-cell coordination features
or higher-order spatial multiplexing can be applied (which
would not be required or provide additional benefit for e.g. less
capacity-demanding cells).

Where is a solution most useful?

No change
Ease-of-deployment, cost and energy efficiency
Coverage
Capacity and performance
Advanced MIMO for

extreme performance
Dense urban high rise Urban Suburban Rural
Different categories of Massive MIMO features may be more nergy efficiency features are applicable to all deployment
E
applicable to some parts of the network than others, depending scenarios as even very high-capacity demanding sites may have
mainly on capacity demand and deployment density. This is a traffic pattern that varies over time (e.g. sites around office
like how different hardware solutions apply to different parts buildings which may be empty during evenings but full during
of the network as discussed in [Ch. 3.2, p. 49-51]. Capacity and daytime).
performance features are generally useful for a large part of
the network, except for less capacity-demanding rural sites.
Coverage enhancement features are most useful in rural and
suburban deployments with larger inter-site distances but are
still needed for dense urban deployments due to e.g. deep indoor
users. The Advanced MIMO features category targets specific
deployments with very high-capacity requirements, such as
dense urban deployments with short ISD, or capacity hotspots
such as stadiums, shopping centers, or trains stations.

Feature evolution of Ericsson

Massive MIMO N
5G coverage, Enablers for improved DL/UL capacity and Efficient and

ease-of-deployment DL/UL capacity. coverage optimizations effective. Massive
and peak rate Further improved peak for high-load networks MIMO adapted to
rate and coverage any conditions
Cost and energy efficiency
Advanced MIMO for extreme performance
Ease-of-deployment
Coverage
Time
As the different categories of software features address different FDD Massive MIMO is typically replacing existing equipment on
network requirements, it makes sense to deploy features bands that have been in service for a long time. Hence, the main
step-wise depending on which network need is most urgent at motivation to deploy FDD Massive MIMO is often capacity. This
different times of the network evolution. Although all software essentially means that FDD Massive MIMO enters the evolution
categories are of importance for all the Massive MIMO segments, path at a later point in time. Coverage and ease-of-deployment
FDD, TDD mid-band and mmWave, the evolution steps and/or must be catered for already from the start to match existing
the focus may be different. deployments.
Taking TDD mid-band as an example. When first rolling out the

mid-band spectrum, focus should be on ease-of-deployment
features and coverage to make sure the Massive MIMO radios
can be deployed in various conditions and that the mid-band
coverage can be extended so that as many users as possible
can access the mid-band spectrum. After the initial deployment
phase, f ocus should shift to features that improve single-user
performance and throughput, and as the traffic network load
gradually increases on mid-band, capacity features need to be
introduced in order to cater for that demand, eventually shifting
focus to more advanced MIMO features which can provide
extreme capacity and performance for high-load scenarios.

Coverage solutions
Limited
change
Access coverage Ways to improve coverage

• Initial access • Improving SNR
• Control channels • Reducing interference (improving SINR)
Data channel (App) coverage • Improving detection reliability given an SNR
• Data channels • Frequency interworking
User performance Limited channel information restricted beamforming
Baseline (access + data)

User Throughput
Single
Improved access wide beam
Improved access + data
SNR Wide beam Restricted beamforming Fully flexible beamforming

Access limit
For coverage enhancements, it is important to differentiate Massive MIMO with beamforming can improve the signal-to-
between access coverage and data channel coverage. Access interference-and-noise-ratio (SINR) on the receiver side, hence
coverage is coverage of the cell-defining signals and messages increase the coverage. A Massive MIMO radio has the possibility
in the initial access procedure as well as the control channel to use beamforming to create narrow beams that maximize
coverage, which is the coverage of the supporting control the beamforming gain in a certain direction, but it also has the
channels needed for scheduling and providing feedback possibility to create wider beams (and thus lower antenna gain).
for the data channels, while data channel coverage is The benefit with using wider beams is that fewer beams are
the coverage of the data channel itself. Access coverage required to cover the entire intended cell service area, which in
should be at least as good as data channel coverage. However, some cases may be desirable. On the other hand, the downside is
having access coverage that surpasses data channel coverage that the experienced beamforming gain is lower.
does not necessarily result in actual improved downlink or uplink
coverage, but rather, it means that some radio network resources Using fully flexible beamforming, with many narrow candidate
are wasted. beams covering the entire cell service area, whenever possible
is beneficial in both downlink and uplink in order to maximize
Different channels and signals have different requirements on coverage. However, there are some exceptions where restricted
signal quality and area coverage. For example, cell-defining beamforming or even a single wide beam may be unavoidable
signals, e.g. SSB, need to have coverage in the entire cell service or more suitable, e.g. at initial access, when there is no or
area, whereas UE-specific channels only need to cover the limited channel information available or when additional
intended UE. This impacts the coverage solution. coverage improvement is not necessary. Good enough channel
information is key to fully reap the benefits of Massive MIMO.
The Massive MIMO architecture can also have a profound

impact on the solution. For example, analog implementations
used typically both at gNB and UE for mmWave require special
feature solutions. This is further discussed in [TP8].

Capacity solutions
In which scenario is each scheme Overview of features and benefit for DL data channel
most useful?
Codebook Codebook
Coverage Coverage
based Capacity Capacity
based
Peak rate Peak rate
Dynamic codebook/reciprocity
Reciprocity- and SU/MU switching
Traffic demand
based Coverage
MU-MIMO Codebook-based
Capacity
MU-MIMO
Peak rate
Reciprocity Reciprocity
Reciprocity-based Codebook-based
based based
SU-MIMO SU-MIMO Coverage Coverage
Capacity Capacity
Peak rate Peak rate
Good coverage Poor coverage
On a high level, the basic features for downlink performance There are different options for how reciprocity-based and
and capacity in a Massive MIMO solution can be classified firstly codebook-based SU/MU-MIMO can be implemented, both from
according to whether codebook-based or reciprocity-based what is available in the 3GPP standard and from proprietary
channel information is used and secondly whether single-user or algorithm perspective. On a high-level, however, the various
multi-user MIMO is applied. This results in four distinct high level options still share the same characteristics with respect to
downlink Massive MIMO feature categories. use case and performance. By comparing the four Massive
MIMO features with respect to the network key performance
There are many aspects that impact which features that are indicators (KPIs) of interest, coverage, capacity and single-user
applicable, and how they are implemented. For example, peak rate, they exhibit different strengths and weaknesses.
reciprocity-based solutions can typically not be used for FDD Codebook-based beamforming has an advantage in coverage
bands, different radio access technologies may need different over reciprocity-based beamforming. Similarly, SU-MIMO has a
feature sets, different UE categories have different constraints, coverage advantage over MU-MIMO as the available transmit
etc. This means that the network must be very flexible and power needs to be split between multiple users in the latter case.
support a wide range of features. Then, which features that are When it comes to capacity, reciprocity-based MU-MIMO has
used at a certain point in time depend on aspects mentioned the highest potential. However, as the coverage of MU-MIMO is
above, as well as traffic conditions, deployed bands and radios, limited, the maximum benefit only occurs in dense deployments
etc. with small cell sizes so that most of the users in the cell have
coverage. To fully utilize the potential of a Massive MIMO
solution, it is necessary to dynamically switch between the four
above schemes so that coverage, capacity and peak rate jointly
can be maximized, which is also how the typical Massive MIMO
solutions are designed.
xamples on acheivable performance for MU-MIMO

E
with Ericsson interference sensing and reciprocity-based
beamforming versus codebook-based beamforming are
presented in [TP9].

Advanced MIMO for extreme performance

No chan
High-order MU-MIMO Cell-shaping

Traffic per cell
5% Coordinated multi-point
20%
50% Extreme capacity demand
(>8 layers MU-MIMO)
Cells
100% High-capacity demand
(≤ 8 layers MU-MIMO)
Low-Mid capacity demand
(SU-MIMO)
Coordination
Advanced MIMO features target more specific deployments or nother example of a class of advanced Massive MIMO features
A
scenarios which may have an extreme capacity need, very high is Coordinated Multi-Point (CoMP), wherein multiple nodes,
load or some other special characteristics. on the same or different sites, coordinate their transmission/
reception in order to mitigate interference or boost the signal
or instance, while MU-MIMO is a key feature to provide capacity
F energy.
in a Massive MIMO solution, the number of MU-MIMO layers
required to fulfill the capacity demand is generally quite low in Another example is cell shaping that can be used to solve various
practice. In fact, in a large portion of the network, the capacity problems. For example, the cell-border with its interference
demand can be fulfilled with only SU-MIMO. Even for cells with impairments can be moved to avoid being in the middle of
high-capacity demand, it is generally sufficient with 8 downlink a location where there are typically many simultaneously
layers and 4 uplink layers to get most of the MU-MIMO gains connected UEs, known as a hotspot, as illustrated in the
with MBB traffic. However, the high-capacity Massive MIMO upper-right figure. An example of a hotspot is a train station. Cell
radios are dimensioned to support 16 downlink layers and 8 shaping also offers the possibility for load balancing among cells
uplink layers, since this could be required in certain cell sites with by moving UEs from one cell to another by altering the uptake
extreme capacity demands, such as traffic hotspots or stadium/ areas of cells.
arena deployments with a lot of people in a small area. The
benefits of very many layers are substantial when they can be or further details on the features see [TP7.3].For illustration of
F
used. The conditions for when they can be used are however very performance, see [TP9].
specific and rarely occurring in reality.

Ease-of-deployment,
cost and energy efficiency No
Directional interference Electromagnetic field (EMF) Energy efficiency Massive MIMO

mitigation exposure handling features
Booster Carrier Massive

sleep MIMO sleep
Mic
ro/
Frequency
Nullforming M illi S
le ep T
x
User data booster Carrier

Traffic booster Carrier
Basic Carrier
Ease-of-deployment features can be used to facilitate assive MIMO solution provides high capacity and performance
M
deployments in some specific cases. which is useful for handling high traffic load scenarios such
as busy hours and occasional peak loads. In most traffic load
All regulatory requirements need to be fulfilled and it is scenarios, even in the most loaded areas, there are still periods
imperative to have features that ensures this. One example is with low traffic load where the available capacity is not required.
that some TDD mid-bands in certain regions have spectrum During these periods, power can be saved by deactivating
coexistence requirements with regulatory limitations on capacity that is not needed. By switching off radio unit
generated interference towards other users of the same components, power consumption is reduced.
spectrum, such as satellite earth station receivers, which may be
operating either in-band or on adjacent band.Another example
is requirements on electromagnetic field (EMF) emissions, which
may require dynamic adaptation of the power level depending
on the beamforming characteristics of the cell.

FDD features for existing bands
LTE Interference rejection SU-MIMO Digital sectorization

combining
NR Interference rejection SU-MIMO MU-MIMO

combining
Although FDD and TDD mid-band features are very similar, Digital sectorization basically splits an existing sector into two
there are some downlink aspects that differ. or more narrower sectors, while keeping the same combined
coverage area as the original sector. The main challenge with
Compared to TDD, FDD cannot straightforwardly use reciprocity digital sectorization is how to map the few cell-specific reference
to acquire detailed channel state information (CSI), as FDD signals (CRSs) to the radio (and antenna) branches in order to
uses different frequencies for uplink and downlink. Therefore, create sectors with good enough isolation and limited inter-site
FDD relies on codebook-based beamforming. The LTE and NR interference.
standard support for codebook-based beamforming is very
good, both for SU-MIMO and MU-MIMO. A drawback with From an uplink perspective, FDD Massive MIMO can offer
codebook-based beamforming compared to reciprocity-based substantially improved coverage, user throughput and capacity
beamforming is a stronger dependency to UE capabilities for both for NR and LTE. FDD and TDD mid-band share many of
codebook-based beamforming, e.g. how many CSI-RS ports and the available Massive MIMO features, especially for the next
type of codebooks a UE supports. generation eCPRI-based FDD Massive MIMO implementation.
Uplink interference rejection combining is one example of a
FDD Massive MIMO is typically deployed in bands with existing Massive MIMO feature available both for FDD and TDD that can
services and therefore need to cater for legacy. For example, boost the uplink performance substantially.
FDD Massive MIMO needs to support LTE. There are essentially
two tracks that can be pursued to improve performance of LTE FDD Massive MIMO can offer improved coverage, user
FDD Massive MIMO deployments, either digital sectorization or throughput and capacity from a downlink perspective as well.
codebook-based CSI using standardized transmission modes. For NR, features such as codebook-based SU-MIMO and
The latter track has generally a better performance potential MU-MIMO using type I and type II codebooks have the potential
but suffers from a strong dependency on the UE capabilities, for to substantially improve performance (see [TP9]). For LTE,
example, which transmission modes (TMs) that are supported. features such as SU-MIMO together with digital sectorization
Unfortunately, most LTE UEs do not support advanced TMs and can boost the network performance significantly.
digital sectorization remains the only viable option.

Beam management for analog

wideband mmWave implementations
Wide beam sweep
Narrow beam sweep
Analog beamforming on network and UE

side for both uplink and downlink UE beam sweep
As discussed in [Ch. 2, p. 35], a competitive mmWave solution One shortcoming of time-domain beamforming using narrow
needs to support very large bandwidths and provide good and high-gain beams is that transmission/reception can only
coverage. Since both wave propagation properties and hardware take place in one direction (one precoder) per beamformer and
technology performance depend on frequency, particularly the time instance. This means that only a limited part of the channel
coverage on mmWave is significantly lower, see [TP8], some can be observed per beamformer and time instance, hence
additional features are required for efficient use of mmWave limited observability, both at network and UE side. To achieve
in addition to the common features outlined above. The and maintain matching beams/directions between the network
purpose of the major part of these features is to adapt to the and the UE, a new beam management framework has been
implementation conditions, to enhance coverage or to cooperate introduced.
with lower bands.
Although many features are similar for mmWave and mid-
Improved coverage is primarily achieved by increasing the band TDD, mmWave solutions need to cater for time-domain
antenna area, both at the network and the UE side. To efficiently beamforming on both network and UE side using beam
increase the antenna area at mmWave with short wavelengths management, which can be described in terms of three
requires typically many radio chains. Furthermore, high antenna procedures. In a first procedure, typically used in the initial
gain means that the corresponding beams become very narrow. access phase, the entire cell is swept with a set of rather wide
network beams so that an initial beam-pair between the network
To support the combination of large bandwidths and large and the UE is established based on the best wide beam. In a next,
antenna gain with reasonable complexity taking implementation second procedure, the initial beam-pair is refined by sweeping a
constraints, interface limitations, heat dissipation, cost, etc. set of narrower network beams within the wide beam coverage
into consideration, an analog architecture using time-domain area. The refined beam-pair link will thus have better coverage.
beamforming has been adopted for the first mmW products. The In a third final procedure the UE can refine its beam, thereby
need for beamforming both at network and UE side based on an further increase the beam-pair link quality. Sometimes the three
analog architecture with time-domain beamforming has some procedures are referred to as P1, P2 and P3.
implications on the realization of mmWave solutions.
For more detailed feature descriptions and 3GPP defined tools to
align transmit and receive beams, see [TP8].


—summary

Summary
• In the Massive MIMO radio, the beam- Hardware features

forming processing is integrated with the
antenna to improve performance and
reduce fronthaul capacity needs
• The Ericsson Massive MIMO radio

portfolio is segmented to address
Software features
different site requirements in terms Massive MIMO
of capacity, coverage and ease-of- mmWave
deployment
Remote radio
• The Ericsson Massive MIMO feature unit (RRU)
portfolio is segmented to address
requirements relating to capacity,
coverage, extreme performance
requirements and ease-of-deployment,
cost- and energy efficiency RAN Compute Passive Interleaved Massive MIMO Massive MIMO
—baseband unit antenna / Hybrid FDD TDD mid-band
The Massive MIMO product architecture is different from he feature portfolio is segmented into four classes, viz.
T
conventional remote radio units (RRUs), also referred to as coverage, capacity and performance, advanced MIMO for
remote radios. Some baseband functionality is moved from extreme performance, and ease-of-deployment, cost and energy
the baseband unit to the Massive MIMO radio to reduce the efficiency, to address the same requirements and constraints
requirements to transfer very large amounts of data between the as the radio products. The segmentation of the features is
Massive MIMO radio and the baseband and to improve latency. somewhat different than the segmentation of the radio portfolio,
since features can be deployed on any site and at any time and
The radio portfolio is divided into three product segments, independent of each other.
capacity, compact and coverage, to meet all performance
requirements and constraints. The capacity segment targets See [14] for a general introduction to Ericsson Massive MIMO.
the most capacity demanding sites and provides superior
performance in all deployment ranging from dense urban high-
rise to rural. The coverage targets environments with larger inter-
site distance like suburban or rural and has high EIRP to ensure
coverage, but typically fewer radio chains, and therefore lower
TCO, compared to the Capacity segment. The compact targets
sites where ease-of-deployment is important. It prioritizes TCO
and ensures that mechanical properties, e.g. size and weight,
comply with site constraints.

4. Guiding principles
for how to select a
radio solution
4. Guiding principles for how to select a radio solution

Ericsson | The Massive MIMO handbook 4. Guiding principles for how to select a radio solution 73
Outline — guiding principles on how to

select a radio solution N
• Steps in the guiding principles

and details of each step Traffic load
• Guide for how to find cost-

efficient solutions for different A
network conditions
B
• Applying the guide C D
E
Capacity
Massive Compact
MIMO
Remote Coverage
radio
Sites
The purpose of this chapter is to provide guiding principles for

how to select cost-efficient radio solutions that meet the network
requirements for typical site conditions. The guiding principles are
built on four network components: strategy, assets, analysis and
solutions. The process and the steps are first outlined. The details
in each step are elaborated. Finally, examples of how the guiding
principles are applied to typical site deployments are provided.

Guiding principles on how to select

aGuiding
radio principles
solutionon how to select a owner
Radio so
radio solution
1.Strategy 2. Assets 3. Analysis 4. Solution
● Services to offer ● Spectrum Needs Radio solution

● Capacity to be ● Sites ● Coverage ● Spectrum
provided
● Equipment ● Capacity ● Massive MIMO or remote
● CAPEX and radio
● User throughput
OPEX budgets
● Array capabilities
Constraints
Feature solution
● Ease-of-deployment
● Coverage
● Total cost
● Capacity
of ownership
● Ease-of-deployment
● Energy efficiency
The steps in the guiding principles are here outlined to show The steps to decide on a network solution are here summarized
why they are needed and how they fit together. As the service before giving some examples on how to apply the process.
provider sets their own strategy objectives, some assumptions The steps outlined in the previous slides should only be seen
are made on what is typically stipulated with respect to network as a ‘guide’. The actual process used is of course decided by the
performance, ease-of-deployment and cost efficiency. These network owner, usually the service provider.
assumptions serve solely to illustrate how the guide can be
applied.
1. It is assumed that the service provider has developed

a network strategy for how to achieve their business
objectives and what their mobile network shall do.
2. It is also assumed that the network strategy includes how

to make use of the available network assets, including
spectrum, sites and equipment.
3. Next, a network analysis is performed to find the

performance requirements and constraints relevant for each
site of the network over the investment cycle.
4. Finally, network solutions are made based on the available

radio and feature portfolio that meet the performance
requirements and deployment constraints from the analysis
in step 3.
Details of each step are elaborated on the following pages.

The strategy defines a service provider’s

long-term
The strategynetwork objectives
defines a service provider’s N
longterm network objectives

1. Strategy 5G willexperience
5G will offer a new offer a new experience
• Services to offer Use case evolution 5G expectations
and where to offer Use case evolution

Entertainment, Instant response time
5G expectations
gaming and AR/VR Universal
• Capacity and user Entertainment, gamingFixed Wireless Access
5G Any application, any
experience Instant response time
time, anywhere
experience to be and AR/VR / Smart home
Universal
Any application, any
Capacity to manage
provided 5Gtransport
Fixed Wireless Access Automotive, experience time, anywhere
traffic growth
/ Smart home and logistics
• How to use network Capacity to manage
and financial assets Automotive, transport traffic growth
and logistics
What ultimately decides how a service provider will evolve ther relevant examples of what could be contained in the
O
their network is the overall objectives of the service provider’s business strategy is the use of different RATs (LTE, NR) and
business and the strategy for how to reach these objectives. different frequency bands. In most cases the service provider
Objectives and strategies can vary considerably. As part of is expected to provide ubiquitous coverage within a certain
these objectives is the position on the markets and how the geographical area and sufficient capacity where and when
service provider wants to be positioned in relation to other needed within that area. The strategy may also provide guidance
service providers in these markets, e.g. to be the leader or a concerning which parts of the network are of strategic value, i.e.
follower. The strategy for how to reach these objectives is the sites identified as “most valuable sites”, and hence where the
expected to include what services to offer, e.g. MBB, FWA, what performance should not be compromised. Another important
service quality to offer, e.g. best in the market, and where in the guidance from the network strategy concerns the investment
network to offer these services, e.g. in cities over a certain size horizon, e.g. whether the network expansion is expected to last
or essentially everywhere, and which price to offer. The strategy for a full investment period of 5-7 years or whether the plan is
has implications on the network requirements in technical to make a radio site deployment for a shorter period, e.g. 2-3
terms, i.e. coverage, capacity and user throughput, as outlined years for capacity upgrade, and then re-evaluated. This will
in Chapter 2. What also is expected from the strategy is how the again guide the selection of radio solution for site expansion.
service provider shall use its existing network, e.g. spectrum, sites Practical aspects relating to network management, e.g. to reduce
and existing equipment, and financial assets, e.g. CAPEX and the number of site configurations may also be reflected in the
OPEX budgets, and hence the financial limits for the network strategy.
investments.

Best use of network assets

Best use of network assets In pdf: w
remove
2. Assets
• Spectrum
• Sites
5G mmW
• Equipment
B42/B43
3500MHz
B7/B38/B41
2600MHz Units on site
ESS B1 2100MHz antennas
NR+LTE dual
5G connectivity B3 1800MHz diplexer
4G NR+NR carrier B8 900MHz filters
aggregation B20 800MHz
3G
B28 700MHz tower mounted
2G amplifier
time radios
The most important radio network assets, i.e. the spectrum Radio sites: Radio sites are often difficult and expensive
assets, the radio sites and site equipment, are crucial for several to acquire and to keep. Service providers typically want to
reasons.Firstly, they are the tools and resources by which the maximize the use of the available sites before acquiring new
service provider delivers the service to the end users. Secondly, sites.
they are also important financially, as they are tangible assets
that carry an actual value. A very common situation today is that the radio sites have been
sold to infrastructure companies and then leased back from
Spectrum assets: Spectrum for all technologies (2G-5G) is the service provider. This is often financially attractive, and it
typically expensive to acquire. The license to use spectrum is an also separates the core business of the service provider and the
investment the service provider must make, and the investment expertise from companies that focus on site infrastructure. In this
is expected to pay off over the licensing period. Hence, the new scenario, site OPEX is closely related to site lease contract
usage of available spectrum is very important for all service agreements.
providers. The existing bands, prior to launching 5G, are usually
occupied for existing radio technologies 2G-4G. Over time, Radio site equipment: On existing radio sites, there are typically
these bands are gradually migrated to the latest technology, i.e. already equipment delivering services on 2G, 3G and 4G. The
5G, to be used in the most efficient way. With 5G introduction, number of frequencies used on sites is driven from capacity
new frequency bands become available both in mid-band requirements. For sites in high- and medium-traffic segments
TDD and in the high-band spectrum (mmWave). The new 5G there are often up to 5-7 bands in operation already. Every band
spectrum is often in the range of 100 MHz, which in many cases usually needs a dedicated radio. Therefore, for many sites where
will add a large capacity potential when used on site. there are high-capacity demands, many radio units are needed.
These radios are often connected to advanced multiband
antennas. When deploying new equipment, it is often a good
opportunity to also consider re-engineering the whole site and
modernize existing equipment. Factors that affect this decision
include the network requirements and constraints, but also the
new product(s) to be deployed, the age of existing equipment
and current lease conditions.

Network analysis to define site needs

N
3. Analysis
• Network load and traffic growth

Traffic
• Load distribution in the network load
Time to content
• Service quality and evolution
• Deployment environment and
performance requirements
All coverage areas ranked on worst-case time to content
• Site build constraints
User experience:
• Total cost of ownership Sites Great Good Fair Poor
Having considered the overall strategy with 5G deployment and ervice quality: Time-to-content is usually a good measure of
S
the network assets, the next step is to gather and make use of service quality, [Ch. 2, p. 24]. Long time-to-content is normally
information from the existing network. From such information, an indication of network congestion and that the site capacity is
it is possible to assess the current load on each site and to predict too low. This may be observed in highly loaded sites during busy
what capacity is needed during the economic lifetime of a new hour. Quality during peak hours requires special consideration,
investment, and specifically where and when the highest load particularly in markets where this is an important differentiator
levels occur. Some key considerations are discussed below. between service providers. It is also often the case that the
requirements on time-to-content evolve over time.
Network load and traffic growth: The current network load and
the predicted growth provide input on the general growth rate eployment environment and network performance
D
over time that the service provider must plan to handle. requirements: Whether the considered site is deployed in
dense urban, urban, suburban or rural environment is important
Load distribution: As explained in Chapter 2 [Ch. 2, p. 29], information
the traffic growth and load per site is unevenly distributed, and to understand how to best evolve it.
thus some sites are highly loaded and require high-capacity
solutions, while other sites are less loaded and hence low- ite build constraints: The actual conditions on the sites are
S
capacity solutions may be sufficient. The high-capacity solutions often setting constraints for how site expansion can be done.
are often more highly valued, since they carry a substantial Some sites in urban areas may be using equipment on all
amount of the total traffic. frequency bands. It may therefore be difficult and/or expensive
to add more equipment on-site. In other areas, there may be
restrictions on visual impact. For example, there may be cultural
buildings or objects which should not be impacted or there may
be esthetic reasons for why antennas should not be notably
visible.
CO: Last, but certainly not least, considerations should be

T
made concerning cost efficiency, or as suggested here TCO
per capacity. Almost all service providers have requirements
on return on investments. Hence, the benefits of the product
deployed must motivate the costs spent. Exactly which
requirements apply and how they are calculated are decided by
the service provider.

Select radio solution to meet the needs
4. Solution
• Conventional solution
Categories of Massive MIMO features
or Massive MIMO
• Product variants Advanced
MIMO for
Coverage
• Software features Cross-layer
extreme
performance
optimization
with other
RAN features
Ease-of-
Capacity and deployment,
performance cost and
energy
efficiency
The last step in the process is to choose the site solution from the the demand of certain functionality in relation to the network
available product portfolio* that meets the requirements and demand. As an example, the initial deployment will not need
constraints over the investment cycle following the analysis in all software features for capacity enhancements since the pure
step 3. addition of new bandwidth will be sufficient in the initial phase.
However, after some years it will become very important to have
The radio portfolio (2023Q1) is illustrated in [Ch. 3.2, p. 56] and the ability to add more capacity by simply activate new features
it contains various Massive MIMO and Remote Radio products to add capacity when needed.
from the segments, capacity, coverage and compact.
*) Availability of different products on different bands will change
The software portfolio, illustrated in [Ch. 3.3, p. 61], contains over time, as market requirements develop. Since the evolution of
a rich set of features addressing the different requirements the portfolio is going very quickly, it is however recommended to
and constraints derived in step 3 of the process. In contrast to consider the products mentioned in this handbook as examples
the radio solution, feature solutions can be implemented more only and always consult the latest available portfolio
gradually as the requirements emerge. A radio product that is
installed year one can thereby be improved in terms of capacity,
coverage and quality some years later when additional software
features are released, and new functionality has become
available in the handsets. The important message here is to see

Case studies—radio solution selection

N
Traffic load
B D
C
E
Capacity
Compact
Massive MIMO
Remote
radio unit Coverage
Sites
Over the following pages, the process outlined above is • The distribution of load and the classification into high-
illustrated with five examples for TDD mid-band, one example (RED), mid- (BLUE) and low-traffic (GREEN) sites have been
for FDD and one for mmWave. To apply the process, the solutions established.
are chosen according to the scenario assumptions made.
• Predictions of YoY traffic growth have been made.
Assumptions:
4. Product portfolio
1. Strategy
• The Ericsson radio and feature portfolio as of 2021H2.
• The service provider offers a 5G user experience in a certain Site A: High-traffic (RED), Dense Urban. Priority: High. ISD:
area, viz. urban and suburban areas with more than 1 000 ~200m. Distribution of building heights: High.
subscribers/km2.
Site B: Medium-traffic (BLUE) Urban. Priority: Medium. ISD:
• The service provider has limited CAPEX and OPEX and intends
~350m. Distribution of building heights: Low. Candidate site for
to make TCO-per-bit-efficient product decisions for site
sharing equipment with other service provider.
expansions.
• A selected number (200) of sites are deemed important, e.g. Site C: High-traffic (RED) Suburban. Priority: High. ISD:
government buildings, stadiums, important businesses. For ~800m. Distribution of building heights: Low. Site D: Low-
these sites, performance has highest priority. traffic (GREEN) Dense Urban. Priority: Medium. ISD: ~300m.
Distribution of building heights: Medium.
• The service provider has an agreement with another service
provider to share RAN equipment where possible. Site E: Low-traffic (GREEN), Suburban. Priority: Low. ISD:
~600m. Distribution of building heights: Low.
2. Network assets
The most effective way to offer 5G user experiences with
• The service provider has acquired 100 MHz mid-band spectrum corresponding user throughput and capacity requirements in
on 3.5 GHz. the whole 5G area is to make use of the available 5G spectrum:
100 MHz at 3.5 GHz on all sites. 100 MHz is a significant chunk of
• The considered network contains all deployment scenario
spectrum that will enable a significantly higher user throughput
types, dense urban, urban, suburban and rural and is covered
and capacity compared to the available spectrum for 4G.
by 2G and 4G with 2000 sites.
3. Deployment considerations
• Roughly 1000 sites, spread over several areas, have restrictions

on deployment.

Case study — Mid-band TDD radio

solution selection
No chan
Traffic load
B
AIR 6419 AIR 3258
Massive Capacity
MIMO Compact
Remote Coverage
radio
Sites
ite A: The location is in an area with high traffic load and high
S ite B: Similar conditions as for site A in terms of traffic and
S
traffic growth. This strongly suggests a product in the Massive traffic growth. The service provider has an agreement to build
MIMO Capacity segment. Since the priority is high, it is a shared network with another service provider, hence the
recommended to unlock the full potential of the mid-band by Massive MIMO product needs to support larger instantaneous
using superior Massive MIMO products and software features. bandwidths than in other areas. Therefore, another product in
Since the distribution of building heights is high, high steerability the Capacity segment with a large IBW is preferred, so that it
in the vertical dimension is needed, thus a 64T64R model. For can accommodate both service providers. As the building height
this specific site, it is assumed that there are no deployment distribution is much smaller, the benefits of a 64T64R product
restrictions with respect to size, weight, wind load or visual is not as significant, and hence a 32T32R product will have
impact. The preferred product choice (2021H2) is the AIR 6419. similar capacity performance but lower TCO. This motivates the
From feature perspective, the available capacity enhancing selection (2021H2) of AIR 3258.
features should be deployed.

Case study—Mid-band TDD radio

solution selection cont.
N
Traffic load
Radio 8863
C D
AIR 3219
E Capacity
Compact
Massive
MIMO
Remote Coverage
radio Radio 4422
Sites
Site C: The location is in a suburban area with rather large Site E: The location is in a low-traffic area and the traffic growth
ISD, high traffic load and high expectations on traffic is low. The site location is on rooftop in sub-urban region and the
growth. Therefore, products in both the Capacity and priority is to expand the site with as little site visual impact as
the Coverage segment can be considered. The distribution possible (small antenna and small remote radio). This suggests
of building heights is low and therefore 32T32R models are that a product in the Compact segment would be suitable.
expected to provide similar performance as 64T64R. The TCO As the current traffic load and the expected traffic growth are
analysis will therefore be useful to guide further concerning low, a Remote Radio would provide sufficient capacity for the
what choice to make. Considering the service provider strategy whole investment period (5-7 years). From a TCO perspective,
to prioritize dense urban and urban deployments, a product in a lower priced solution is preferred (lower price on remote radio,
the Coverage segment is chosen as the TCO is lower. Among the less capacity demand from baseband licenses, less energy
available products (2021H2) an AIR 3219 is an adequate choice. consumption, less complex and cheaper antenna). In this
example, the Radio 4422 is a suitable choice (2021H2).
Site D: This location is in an urban area with low ISD. The actual
traffic and the expected traffic growth is low. Therefore, the
TCO-per-capacity analysis indicates a benefit for a conventional
solution based on remote radios. The capacity of a remote radio
is considered sufficient. The priority of the site is considered high
due to its location. Therefore, a remote radio in the Capacity
segment using 8T8R is preferred. There are no deployment
restrictions. In this case a reasonable choice (2021H2) is Radio
8863.

Case study—Mid-band TDD network

deployment
AIR 3227
AIR 3268
AIR 3255
Traffic
load
Macro Sector deployment

AIR 6419
Capacity
Compact
Massive
MIMO
Radio 8863
Remote Coverage
radio unit Radio 8823
Sites
This is an example that illustrates the 5G TDD mid-band The high-traffic sites in the main cities and connecting roads
deployment by a European mobile service provider and how were considered high-priority and speed and capacity were
the principles for product selection have been applied to meet prioritized. The operator intended to provide the best end-user
their deployment goals. The operator targets nation-wide 5G service in the market and to make the investments needed to
deployment. They operate in a market where there are three ensure such performance in the high-traffic sites. A high-quality
mobile operators competing on similar terms. service should be provided also in the medium and low traffic
sites in the cities and in the small towns, but with a reasonable
In the network there is a mix between typical European style TCO. To simplify the network roll-out and operation, the number
cities*, European style small towns** and rural areas. of product types should be limited.
The roll-out was based on three goals: The strategy used for achieving nation-wide coverage was to
provide seamless 4G/5G inter-working using Ericsson Spectrum
• To be first in the market with 5G.
Sharing (ESS). To start with, speed and capacity demands in
• Provide the largest 5G coverage. high-profile and high-traffic areas were achieved by deploying
of AIR6488 in main cities and connecting roads. Later, the
• Offer a 5G network with highest user data rates. high-profile, high-traffic areas in the middle-size cities were also
covered. In the rural areas, where the capacity requirements are
lower, Radio 8823/8863 were chosen to fit the purpose.
AIR6488 was later complemented by several versions of

AIR32xx products, which had better capacity-fit in lower-traffic
areas and had lower energy consumption. The Macro Sector
was used to combine AIR32xx with passive antennas for a more
streamlined site appearance.
*) The traffic in the city centers is in the range medium-high

and the buildings are predominantly in the range 4-6 floors.
The number of high-rise buildings is very low.
**) The building heights in the small towns are somewhat lower
than in the cities, typically 2-4 floors.

Case study—Mid-band FDD radio

solution selection
Case study—FDD radio solution selection C
Traffic load
AIR 3283
Capacity
Massive
Compact
MIMO
Remote
radio Coverage
Sites
The main use case for FDD Massive MIMO is to increase capacity As the FDD mid-bands are not yet fully migrated to NR,
of existing FDD bands. It is primarily FDD mid-band macro sites the dominant Massive MIMO feature used in LTE is digital
that get upgraded to Massive MIMO. The FDD Massive MIMO sectorization, splitting the current sectors in either 2 or 3
units currently on the market are dimensioned with respect to narrower sectors. In the parts of the spectrum that are migrated
output power and antenna configuration to match the coverage to NR, more advanced MIMO features can be used thanks to
provided by the existing equipment. more advanced NR UE MIMO capabilities.
The example (see picture) illustrates a site that is deployed in a As the FDD Massive MIMO radio units have multiband capability
typical north American urban area. The existing solution cannot and support mixed operation of LTE and NR in each band, the
provide sufficient capacity and user performance. To improve service provider can gradually migrate from LTE to NR as the UE
the situation (before being able to acquire additional frequency population is maturing.
bands), the service provider is replacing the existing 4T4R
solution with a Massive MIMO 16T16R solution (e.g. AIR 3283)
that in this case improves the offered capacity with about 50%.

Case study — mmWave radio

solution selection
Zooma
Can we
the pictu
https://b
Traffic load
Streetmacro
6705
Capacity
Massive
Compact
MIMO
Remote
radio Coverage
Sites
In most countries, mmWave is today used primarily in special This kind of mmWave deployment offers huge capacity due
deployments with very high capacity needs. to large bandwidths and the special dense deployment makes
coverage good enough. During special events the service
The example here illustrates a sport stadium, where the provider can deploy many mmWave street sites, with roughly
capacity demand during sports events can be very high, or even a ten-fold increased user performance, to provide the required
extremely high, with about 10-100 times the capacity per m2 capacity.
compared to a typical dense urban deployment. Under these
extreme conditions, a dense deployment of mmWave radios, e.g.
Streetmacro 6705 is an attractive option to cater the extreme
capacity demand.

Summary — guiding principles on how

to select a radio solution
No
• To find cost-efficient radio solutions and to 1.Strategy 2. Assets 3. Analysis 4. Solution

find where Massive MIMO provides superior
performance, considerations concerning
network strategy, assets, site-based analysis
and available radio solutions are suggested AIR 6419
A
• Applying the guideline demonstrates what Radio 8863
product solutions provide good performance AIR 3258

B
D
C
and cost-efficiency E
• Examples including mid-band TDD, mid- Massive Capacity

MIMO Compact Radio 4422
band FDD and mmWave have been AIR 3219 Remote Coverage
considered radio
The service provider decides what their network shall deliver to I n order to deliver a 5G experience, the new 5G mid-band
the end users and a strategy for how to fulfill these objectives. spectrum is the solution to improve user experience with
The service provider then needs to find radio solutions for each improved throughput and capacity. In parallel, and/or when
site that meet the network requirements that follow from their TDD mid-band spectrum is not yet available, the service provider
strategy with respect to capacity, user throughput and coverage should maximize the usage of all its spectrum, including legacy
under the constraints of ease-of-deployment, cost efficiency and FDD and mmWave. What products to use for different spectrum
energy efficiency. The solution may involve any means available, bands is then up to the service provider to decide based on the
e.g. taking new spectrum into use, finding a suitable product or guidance in this handbook.
site expansion. As radio sites are expensive to acquire and to
The main steps are:
operate, the preferred solutions build on the existing site grid.
Once these are fully utilized in terms of use of available spectrum, 1. Specify the deployment objectives in different areas based on
equipment and functionality, site expansion can be considered. the network strategy.
2. What spectrum shall be deployed.
3. What are the deployment conditions at each individual site.
4. What products are available.
Selecting a set of products that meets the performance
requirements, considerations need to be taken to deployment
conditions and the requirements on cost efficiency.


5. Future developments
Ericsson | The Massive MIMO handbook 5. Future developments 87
Developing services and technologies
This handbook is focusing on existing wide area networks and Several new services are expected to reach the market in
operator services, and mainly on mobile broadband (MBB). the years to come, e.g. XR, and industry applications. Some
The applicability of the suggested solutions are now and in the of these services are very demanding with respect to user
near-term future. There are however some changes expected to throughput and/or latency. This will drive a need for improved
come in the near-to-mid-term future. Other services than MBB, network performance, which in turn will also push the need for
which has been the dominating mobile service for the last 15 performance enhancing technologies, such as Massive MIMO.
years, which will put new and more advanced requirements on
the network may develop new eco-systems and gain popularity. In addition to new services, new deployments and new
Due to this, at least in part, new deployments may need to be technologies will also be introduced, which may impact the
adopted. New services will demand more of technology and development of Massive MIMO, either in terms of performance
introduce new innovative features. or hardware effectiveness. Most deployments of new services
are expected to be made in public service provider networks,
but there are also expectations on private networks to grow in
numbers.
In this chapter, some promising services, deployments and new

technologies that may come into use in the years to come are
introduced. It should be kept in mind though that the success
and importance of these have for most cases not yet been proven
in real networks.
1. Future
5.1 New developments
and developing
services
Massive MIMO unlocks the full potential

of Fixed Wireless Access (FWA)
Mi
e.g
Beamforming and MU-MIMO opportunities
• Fixed user devices, often in favorable locations, e.g., on

roof tops or close to window
• Form factor which allows high performance antennas
• Long data sessions, e.g. video streaming
• Good signaling conditions allow for mmWave extended
range
FWA: providing fixed broadband over mobile networks

as an alternative to xDSL, cable or fiber networks
Fixed wireless access (FWA) provides fixed broadband FWA networks are therefore a sweet spot for Massive MIMO
connections over mobile networks [3]. The service is an solutions and deployed capacity enhancement investments can
alternative to fiber, cable or arbitrary digital subscriber line quickly be rewarded by larger revenues. For example, advanced
(xDSL) networks and is characterized by fixed user locations, beamforming capabilities improve the link budget and thus the
typically households or small to medium sized enterprises, and spectral efficiency. Outdoor customer premise equipment (CPE)
high data consumption levels. allows for better antenna solutions which in turn make more data
layers possible.
The data consumption levels together with requirements
for sufficiently good rates also for the worst located homes As illustrated in the figure to the right, in a downlink capacity
during peak traffic, put high capacity requirements on the limited FWA scenario, the data is sent to a limited number of
communication service providers’ network dimensioning. stationary CPEs, and as the traffic is dominated by data in longer
Downlink capacity is often the limiting factor in FWA sessions with relatively large packet sizes, FWA is an example of
dimensioning due to that the traffic is dominated by video where MU-MIMO is beneficial.
related content in the downlink. Consequently, any capacity
improving feature is important for FWA. Another way to improve capacity is to use mmWave extended
range where mmWave radios on macro base stations are used
to offload lower bands when so required. Large antennas with
a high degree of beamforming increase the signals levels which
are required to achieve sufficient mmWave range and coverage
[TP8].
Cloud gaming and AR expected

to drive Massive MIMO in public
networks No chan
No chan
Zooma
.. Or will
Cloud gaming and extended reality (XR) applications, in Simple applications—remote rendering
particular augmented reality (AR), have the potential to drive
network requirements far beyond mobile broadband use case Typical characteristics for remote rendering are:
requirements. This is mainly driven by the excessive use of visual
• Downlink bit rate: 5 Mbps for simple remote-rendered content.
content and the potential for immersive communication in areas
such as tele-meeting, shopping, gaming, and entertainment. • Uplink 0.5 Mbps: for pose and control information send to the
remote rendering engine for content generation.
The development of the XR market still lies ahead, but XR is
expected to be the main new user interface replacing the phone • Roundtrip time latency 40 ms (RAN+Core part): low latency
screen. Mass market uptake will require slim form factors, needed for remote rendered content
implying less space for battery and compute. Slim devices
will require compute offload, which further will drive network
capabilities and, in particular, network coverage and capacity. Dynamic application and environment
The user experience will drive requirements on latency as well as • Downlink 100 Mbps: High quality volumetric streaming of
capacity, and where the most basic AR applications are expected content, e.g., 8 M points @ 30 fps => 120 Mbps per object.
to have moderate requirements, advanced use cases will put
• Uplink 50 Mbps: Large sets of high-resolution sensor data
excessive requirements on network. Two examples are next
uploaded to cloud for environmental understanding and correct
given.
rendering. Uplink data for spatial mapping (~10-20 Mbps) +
high resolution 3d point cloud upload for enhanced experience
(~30-40 Mbps).
• Roundtrip time latency 20 ms (RAN+Core part): Cloud based

spatial mapping/Simultaneous localization and mapping.
Network wide application coverage will be challenging

especially in the uplink and use of Massive MIMO, time-critical
communication features as well as stepwise densification,
starting in targeted areas are needed to improve coverage
where most needed.
Industrial digitalization drives use

of Massive MIMO
Industrial digitalization drives use of C
N
Massive MIMO
Immense uplink
● Immense
● uplink video
video in
in surveillance
surveillance and
and remote-
remote-
control drive
control drive uplink
uplink capacity
capacity requirements
requirements
● Nullforming
● Massive MIMOto mitigate inter-cell
to mitigate interference
intercell interference
SU-MIMO and
● SU-MIMO
● and MU-MIMO
MU-MIMO spatial
to further enhance to
multiplexing
capacity
further enhance capacity
Remote control of complex machinery such as ship-to-shore Shifting TDD DL:UL ratio to become more uplink heavy may
cranes (StS) or rubber tire gantry (RTG) cranes drive an excessive mitigate the uplink problems in scenarios where this is possible,
amount of uplink video. For example, an RTG may have up to but probably more likely in high-band than mid-band due to
20 simultaneous high-definition (HD) cameras for the purpose co-existence with public network deployments.
of controlling the crane in different actions. A port could have
tens or even hundred of these cranes. Uplink capacity will put In indoor deployments, using a carefully selected balanced TDD
requirements on Massive MIMO features such as uplink multi- DL:UL ratio may enable sufficient uplink capacity in both mid-
user-MIMO, single user-MIMO, and utilization of mid as well as and high-band. Inter-cell interference is however expected to
high band. limit the gains achievable through densification unless Massive
MIMO with nullforming is used. Finally, cell densification and
the use of Massive MIMO are essential when addressing low
bite-rate TCC services as well as high bite-rate services.
Use of mid- and high-band Massive

MIMO essential in smart port
automation Change
Zooma
text to th
Mid-band capacity per sector High-band capacity per sector

Total traffic served [Gbps] Total traffic served [Gbps]
4 sites 6 sites 8 sites 4 sites 6 sites 8 sites
To meet the connectivity requirements in smart ports of the The figures on the right top shows the capacity of a 5G network
future, 5G New Radio (NR) is the most effective and cost- in a port area in terms of the data volumes served per sector
efficient technology available. In most markets around the in gigabits per second. Owing to the downlink-heavy TDD
world, 5G spectrum is allocated on TDD mid-band, as well as on pattern, downlink capacity significantly exceeds uplink capacity.
mmWave. Likewise, the capacity achieved on the high-band is significantly
superior to mid-band capacity, due to the larger bandwidth.
A port presents significant challenges to the capacity of a Interestingly, the capacity per sector decreases for mid band with
wireless network that mandate a dense deployment of base an increasing number of sectors, while it increases for high-band.
stations. On the other hand, ports resemble a relatively open
propagation environment, which together with the short The reason for this is twofold: first, unlike for mid band, coverage
distances between base stations cause excessive levels of on high-band still improves when increasing the number of
inter-cell interference. Stringent requirements on reliability and sites from four to eight. Second, the larger array dimension in
bounded latency add further challenges to the ability to deliver high-band produces a narrower high-gain beam giving rise to
the required network capacity. an improved signal-to-interference-plus-noise ratio (SINR). As
densification in the high-band improves link budget, it will also
allow a power limited UE to allocate more physical resource
blocks (PRBs) and hence support more data.
For more details on this example, please see [15].
Non-terrestrial network for ubiquitous

coverage with Massive MIMO
Satellite 600 km and above

network High altitude platform station
as base station Air-to-
ground network 20 km
Unmanned aerial
vehicle (UAV) 10 km
(aka. Drone)
150 m
Remote Rural Urban Ground level
One domain of applications, where Massive MIMO will most • Service continuity: Provide continuous 5G services whilst
likely be applied is non-terrestrial networks (NTN), particularly moving between terrestrial and non-terrestrial networks, e.g.
the satellite-driven 5G NTN considered by 3GPP. NTN can airborne and maritime platforms.
refer to a wide variety of solutions, e.g. satellites, high-altitude
platform station (HAPS), air-to-ground (A2G) and unmanned • Service ubiquity: Provide 5G services in unserved or under-
aerial vehicles (UAV). NTN via satellites mainly aims at covering served geographical areas, where terrestrial network may
users in remote and rural areas where no terrestrial network is not be available.
available. • Service scalability: Provide multicast or broadcast of non-time-
sensitive data (e.g. ultra high-definition TV).
The reason for use of 5G non-terrestrial networks is to support
new use cases, and thereby enable new business opportunities, • Resiliency: Enable secure, highly reliable, rapid and resilient
that cannot be supported in a cost-efficient way or not at all from deployment scenarios like emergency response during
ground-based systems, e.g. the following: disasters.
There are several reasons for using Massive MIMO for non-
terrestrial networks, for example:
• Beamforming to provide sufficient coverage and enhance

individual user performance.
• Spatial multiplexing to provide high total network capacity,

as the satellites cover very wide areas.
Note that there is a need to adapt to satellite movements.

For example, the low earth orbit (LEO) satellites move at about
7.5 km/second at 600 km height and will cover a fixed earth cell
for a couple of minutes.
1. Future
5.2 New developments
technologies
6G will have both traditional capabilities

and new capabilities Mo
Security Service
and privacy availability
Data rates
Device Latency Network Service

diversity capacity versatility Inner ring
Outer ring
Cost-efficient of classical
of new dimensions
and sustainable capabilities to be
to be addressed
solutions enhanced in
by networks networks
Positioning Coverage
Network
Network Energy Dependable
sensing performance compute
Deployment
flexibility
Around 2030, the first 6G networks will likely start appear, a The new services and deployments, including a complementary
time at which 5G has been in operation for ten years, shaping satellite component as described in 5.1, are relevant also in the
and transforming the society with new needs and services. 6G context, and on the following slides, some new technologies
Currently, the main drivers for the new technology are the needs candidates applicable also for 6G, are elaborated on.
for trustworthiness, sustainability, accelerated automatization
and digitalization as well as limitless connectivity meeting • For extreme performance and coverage, the spectrum range
the demands for communication anywhere, anytime, and for from sub-GHz to beyond 100-GHz is considered, including new
anything. To meet these future challenges, 6G needs to continue spectrum in the range of 7-20 GHz. For these frequency ranges,
to push beyond the technical limits of 5G, moving toward critical larger antennas with many elements can substantially increase
services, immersive communication, and omnipresent IoT. In system coverage, capacity and user throughput.
addition, entirely new capability dimensions should be explored
• Distributed MIMO, with many coordinated network nodes
integrating compute services and offering functionality beyond
offering connectivity to multiple nodes simultaneously will also
communication such as spatial and timing data such as sensing.
contribute to extreme performance and coverage in term of
Thus, the future platform for a vast range of new and evolving higher data rates, more consistent quality and robustness.
services needs to include both traditional capabilities, as • Artificial intelligence (AI) and machine learning (ML) will be
well a new capabilities. In the figure above, where the target considered for automated operation and optimization of the
capabilities for a 6G network are shown, the center illustrates future radio networks offering a large number of services with
the requirement on not only cost-efficient solutions but also accelerating cost and complexity. Thus AI/ML is not only of
sustainable solutions. The inner ring illustrates traditional interest to Massive MIMO applications, but also to automate,
capabilities such as achievable data rates, latency, and system improve and optimize the networks operation in terms energy
capacity. The outer ring illustrates new capabilities that have efficiency, performance, and to ensure service availability.
typically not been in focus for previous systems.
More reading is available in [16] and [17].
Larger antenna arrays

Main drivers and areas of use
Larger antenna arrays xGHz
Notes pa
Main drivers and areas of use
Today New TDD bands New TDD bands

mid-band Existing antenna è Matching mid-band
Coverage challenging coverage
192 AEs at 7 GHz
192 antenna elements (AEs)

at 3.5 GHz 768 AEs at 7 GHz
Today, Massive MIMO antennas have typically 128 or 192 There are several ways to make use of this fact. When
antenna elements (AEs), which defines the physical size of the introducing new bands, these are almost always on higher
array. As technology to build the arrays advances and the arrays frequencies. In the time frame of 6G introduction, two frequency
become thinner and lighter, there is increased industry discussion ranges, 6-8 GHz and 10-15 GHz are discussed. To make use of
on larger arrays, meaning arrays with more antenna elements, these bands using the existing site grid, the new solution for the
for example 256, 384, 768 and even larger numbers of antenna higher band must address the lower coverage. By increasing the
elements. One main driver for the industry discussion is that antenna area, this coverage difference can be reduced or even
new higher frequency bands are considered for 5G and 6G, for eliminated. So, for example, 64T64R Massive MIMO antenna
example between 6 GHz and 15 GHz. with 192 antenna elements will be physically smaller at 7 GHz
then at 3.5 GHz. The loss in received power due to the shorter
As explained in [TP1], when going to higher frequency bands, wavelength will mean that the antenna coverage is reduced.
coverage is reduced for antennas with a specific number of When the frequency is doubled (and so the wavelength is
antenna elements. At the same time, the physical size of the halved), there is a loss of a factor of 4, i.e. 6 dB, in this example.
antenna at the higher frequency band with the same antenna
gain also decreases, as the size of the antenna elements The coverage increase from larger antennas can also be used
themselves and the gap between antenna elements is for reducing output power or improving performance on existing
proportional to the wavelength*. bands. It is, however, generally believed that the benefits on
existing mid-bands is limited and that greater benefits will be
Fortunately, it is possible to increase the gain of the antenna, seen from coverage extensions on higher bands.
by increasing its physical size. For a Massive MIMO antenna
this means increasing the number of antenna elements. If *) Note that the free-space path loss in the radio channel itself
the antenna area, figure to the left, is doubled the maximum is the same for all frequencies. The difference is due to the
array gain increases by a factor of two (or 3 dB), and if the difference in antenna size.
area is quadrupled the maximum gain increases by 6 dB. If the
frequency is doubled, e.g. from 3.5 GHz to 7 GHz, the antenna
size is reduced to a quarter, see figure in the middle. If the
number of antenna elements is increased a factor four, the
resulting antenna, i.e. figure to the right, will have the same size
and maximum gain as the original one, i.e. figure to the left.
Coordinated multi-point (D-MIMO)

techniques to improve performance
No
Cell Downlink Uplink

shaping coordinated beamforming joint reception
Slow adaptation Fast adaptation Fast adaptation
Coordination Coordination
Reduced inter-cell interference
Improved coverage Reduced inter-cell Improved coverage
interference
Load balancing Reduced inter-cell interference
Coordination between cells for transmission/reception can be Many different flavors of CoMP exists, with different pros and
used to improve network performance, primarily by either aiming cons, different requirements on backhaul/coordination and
to reduce inter-cell interference or utilizing the transmission/ applicability to different scenarios. Schemes include coordinated
reception resources of multiple cells to boost the signal energy of scheduling, coordinated beamforming, dynamic point selection,
a user. and joint transmission/reception. These schemes typically focus
on fast adaptation for data channels.
Such coordination between network nodes or points is referred
to as coordinated multi-point (CoMP) in 3GPP. The term “point” The more advanced forms of cell shaping can also be seen as a
can here for simplicity be thought of as corresponding to the form of CoMP, albeit with a slower adaptation rate. Cell shapes
antenna array of an individual cell. Distributed MIMO (D-MIMO) are then automatically determined based on observations of the
is another, somewhat vaguer, term originally stemming from the environment for example via AI related methods.
academic literature. Most of these techniques are not new but
have for various reasons not yet been adopted in large scale in Due to the large number of digitally controlled radio chains,
commercial systems. In the time frame of 6G, it is expected that Massive MIMO systems have great possibilities to suppress
new use cases and deployments may increase the need for such interference via nullforming using only combinations of signals
solutions. from within a single point. This makes coordinated beamforming
(CBF) a prime candidate for performing coordination in
downlink thereby mitigating interference towards victim UEs in
neighboring cells. For uplink, combining signals between points
is, relatively speaking, easier and less risky than doing the same
in downlink. Coverage is also more important because of the
limited user transmission power. Joint reception is therefore
a promising coordination scheme in uplink that can improve
coverage as well as suppress interference via nulls formed across
points.
1. brings
AI/ML Future developments
improvements in capacity
and ease of operations for Massive MIMO No chang
Capacity and
performance
Advanced MIMO for

extreme performance
Ease-of-deployment,
cost and energy efficiency
Artificial intelligence (AI) and machine learning (ML) For the success of AI/ML it is essential to have access to large
technologies have been successfully applied in various industries, sets of relevant and reliable data representing key dynamics
including radio access networks. The application of AI/ML in live networks. Machine learning models can then be
techniques in mobile networks is rather recent and is expected trained to adapt to real scenarios and to enhance network
to increase in the future. The use of AI/ML is mainly to address performance. The use of AI/ML puts additional requirements
network optimization, e.g. network planning, deployment, on both hardware and software implementation. The hardware
tuning, cost reductions and improved energy performance. must be designed to support the model training and memory
requirements in relation to cost, energy efficiency, footprint, etc.
For Massive MIMO there are already several applications of AI/ while software must be specifically optimized for efficient use of
ML, and the expectation is that the use of AI/ML will increase hardware resources.
further in the future. Examples include elimination of passive
intermodulation (PIM), beam management (particularly for
mmWave), and cell shaping, see Intelligent Cell Shaping [TP7.3,
p. 114]. The ongoing work in 3GPP addressing AI/ML with
specific focus on the physical layer, data collection and model
management is an enabler of a wider application of AI/ML in the
future.
6. Summary
6. Summary
Ericsson | The Massive MIMO handbook 6. Summary 101
Summary
No c
• 5G networks require cost-efficient solutions

that provide significantly improved coverage,
capacity, and user throughput to support
higher capacity and enhanced user Mbps
experience
• All sites are unique with different 750
FDD + NR TDD
performance requirements and deployment FDD bands Conventional
related constraints 500
• Massive MIMO technology improves network 250

performance by using spatial properties of
the radio channel and antenna array 100
Downlink user
capabilities throughput
Traffic demand according to 2025
• Massive MIMO can effectively provide the 50
predictions
required increase of coverage, capacity and
10
user throughput FDD + NR TDD
Massive MIMO
Massive MIMO is a very powerful technology that can enhance Massive MIMO solutions are useful tools for providing the
network coverage, capacity and user throughput by making use network performance improvements, viz. coverage, capacity and
of multi-antenna techniques, such as beamforming, nullforming user throughput, required as the network evolves. The ability to
and spatial multiplexing (MIMO), and antenna array properties. increase coverage on higher bands is particularly important since
this enables the full potential of the new mid-band spectrum on
In an evolving 5G network, there is a need for increased the existing site grid.
coverage, capacity and user throughput to support increasing
network traffic load and expectations on user experience. In Massive MIMO radio- and feature solutions can be designed to
addition to these performance requirements, there are also meet certain performance requirements and deployment related
deployment related constraints for each radio site, e.g. ease-of- constraints to improve relevant performance aspects and to
deployment, cost- and energy efficiency. All sites are unique, increase cost efficiency. Ericsson has designed the radio portfolio
with different requirements and constraints. Massive MIMO is a into three segments, viz. capacity, coverage and compact to
versatile technology addressing many requirements. address the specific requirements for high capacity, coverage in
wide areas and ease-of-deployment. The Ericsson features are
segmented into feature classes, capacity, coverage, advanced
performance and ease-of-deployment, cost- and energy
efficiency.
6. Summary
7. Abbreviations
Ericsson | The Massive MIMO handbook 7. Abbreviations 103
3GPP 3rd generation partnership project DL downlink
2D two dimensional DMRS demodulation reference signal
A2G air-to-ground EIRP equivalent isotropic radiated power
AAS advanced antenna systems; EIS equivalent isotropic sensitivity

note: synonymous to Massive
MIMO EM electro magnetic
ACK acknowledgement (positive) EMF electro magnetic field
AI artificial intelligence EPO energy performance optimizer
AIR antenna integrated radio FCC federal communications commission
AOSA array of subarrays FDD frequency division duplex
API application programming FH frequency hopping

interface
GHz giga Hertz
AR augmented reality
GP guard period
AS antenna switching
gNB node B (NR)
BF beamforming
HARQ hybrid automatic repeat request
BB baseband
HAPS High-altitude platform
BW bandwidth
HW hardware
CA carrier aggregation
Hz hertz
CAPEX capital expenditures
IBW instantaneous bandwidth
CBF coordinated beamforming
ICNIRP International Commission on Non-
CC component carrier Ionizing Radiation Protection
CCH control channel IEEE institute for electrical and

electronics engineers
CE Channel element
IRC interference rejection combining
CO2 carbon dioxide
ISD inter-site distance
CoMP coordinated multi-point
KPI key performance indicator
CQI channel quality indicator
MAC medium access control
CPE customer premise equipment
MBB mobile broadband
CSP communication service provider
Mbps mega bit per second
CSI channel state information
MIMO multiple input multiple output
CSI-RS CSI-reference signal
ML machine learning
DC dual connectivity
mmWave millimeter wave
DCI downlink control information
MCG master cell group
DFT discrete Fourier transform
7. Abbreviations
MSGx message x RAN4 3GPP working group for radio

performance and protocol aspects
MU-MIMO multi-user MIMO
RAN5 3GPP working group for mobile terminal
NACK negative ACK conformance testing
NR new radio RAR random access response
NTN Non-terrestrial networks RMSI remaining system information (SIB1)
OPEX operational expenditures RI rank indicator
ORAN open RAN architecture RRC radio resource control
OSI open systems interconnection RF radio frequency
PA power amplifier RTG rubber tire gantry
PBCH physical broadcast channel SCell secondary cell
PCell primary cell SCG secondary cell group
PDCCH physical downlink control channel SCS subcarrier spacing
PDSCH physical downlink shared channel SIB x system information block x
PIM passive intermodulation SINR signal-to-noise-and-interference ratio
PMI precoding matrix indicator SNR signal-to-noise ratio
PRACH physical random access channel SON self-organizing network
PUCCH physical uplink control channel SUL supplementary uplink
PUSCH physical uplink shared channel SpCell special cell
PSCell primary SCG cell SR scheduling request
PSK phase shift keying SS synchronization signals
PSS primary synchronization signals SRS sounding reference signal
QAM quadrature amplitude modulation SSB block containing SS and PBCH
QPSK quadrature PSK SSS secondary synchronization signal
QoS quality-of-service StS ship-to-shore
Qx quarter x; x=1,2,3,4 SU-MIMO single-user MIMO
R receiver radio chain SW software
RAN radio access network T transmitter radio chain
RAN1 3GPP working group for radio layer 1 TCBW total configured bandwidth
(physical layer)
TCO total cost of ownership
RAN2 3GPP working group for radio layer 2
and radio layer 3 radio resource control TDD time division duplex
RAN3 3GPP working group for UTRAN/E- TRP transmission point

UTRAN/NG-RAN architecture and
related network interfaces
7. Abbreviations
Ericsson | The Massive MIMO handbook 7. Abbreviations 105
TTI transmission time interval
UAV unmanned aerial vehicle
UCI uplink control information
UE user equipment
UL uplink
VR virtual reality
XR extended reality
YoY year-on-year
7. Abbreviations
8. References
Ericsson | The Massive MIMO handbook 8. References 107
1. Advanced Antenna Systems for 5G Network Deployments: [16] 6G – Connecting a cyber-physical world, Ericsson White
Bridging the Gap Between Theory and Practice 1st Edition, Paper, available at: https://www.ericsson.com/en/reports-and-
Elsevier 2020, ISBN: 978-0-12-820046-9, Advanced Antenna papers/white-papers/a-research-outlook-towards-6g
Systems for 5G Network Deployments - 1st Edition (elsevier.
17. [17] Blog “The 6G series: Why it’s a great time to start
com)
talking 6G”, available at: https://www.ericsson.com/en/
2. Advanced Antenna Systems for 5G networks, Ericsson blog/2022/11/why-its-time-to-talk-6g
White Paper, available at: ericsson.com/en/reports-and-
[18] “Ericsson Spectrum Sharing”, available at: https://portfolio.
papers/white-papers/advanced-antenna-systems-for-5g-
ericsson.net/Main-Catalog/Networks/Ericsson-Radio-System/
networks
Radio-System-Software/Ericsson-Spectrum-Sharing/c/
3. Fixed Wireless Access Handbook, EN/LZT 1/28701-FGD FGB101843
101 449 Uen Rev H, Ericsson AB 2022, available at: https://
foryou.ericsson.com/fwa-handbook-2022.html
4. Ericsson Mobility Report Nov, ericsson.com/en/mobility-
report
5. “On the road to breaking the energy curve”, available
at: https://www.ericsson.com/en/about-us/sustainability-
and-corporate-responsibility/environment/product-energy-
performance
6. “Ericsson Silicon- At the Heart of Ericsson Radio System”
https://www.ericsson.com/en/ran/ericsson-silicon
7. Ericsson, IMT-2020 self-evaluation: Radio Network Energy
Performance, 3GPP TSG-RAN WG1 #91, R1-1720954, Reno,
US, November 27-December 1, 2017, available at: https://
www.3gpp.org/ftp/tsg_ran/WG1_RL1/TSGR1_91/Docs/
R1-1720954.zip
8. Ericsson blog, 5G energy consumption: what’s the impact
of 5G NR in real networks?, October 8, 2021, Frenger, P; Jading,
Y; Bengtsson, J, available at: 5G energy consumption: The
impact of 5G NR – Ericsson
9. Blog “Ericsson RAN Compute: Why a smaller footprint
produces a bigger 5G deployment”, available at: https://www.
ericsson.com/en/blog/3/2021/7/ericsson-ran-compute-and-
5g-deployment-insights
10. Blog Ericsson Uplink Booster, available at: ericsson.com/
en/blog/2020/7/uplink-booster-significant-to-the-new-normal
11. Blog Start simple with the Ericsson Radio System,
available at: Start simple with Ericsson Radio System Portfolio
- 3Cs
12. Blog How to build high-performing radios, available at:
ericsson.com/en/blog/2021/1/how-to-build-high-performing-
radios
13. Blog How to build high-performing Massive MIMO
systems, available at: https://ericsson.com/en/blog/2021/2/
how-to-build-high-performing-massive-mimo-systems
[14] “More with less – Introducing ultra-wideband Massive
MIMO radio”, available at: Massive MIMO solutions accelerate
5G mid-band – Ericsson
[15] “Realizing 5G smart-port use cases with a digital twin”,
Ericsson Technology Review, available at: Realizing 5G
connected-port use cases with a digital twin – Er6csson
8. References
Ericsson | The Massive MIMO handbook Overview 109
second edition
Massive MIMO
Handbook
Technology Primer
2 Ericsson | The Massive MIMO Handbook — Technology Primer
Introduction
The purpose of this Technology Primer is to provide a deeper The different chapters of the Technology Primer can be read
understanding to how Massive MIMO works, why it works and selectively and standalone to deepen knowledge where the
what performance is achievable in a real network deployment. reader chooses. The chapters are, however, organized in a way
Many related topics that provide additional insights to the that they best are read in succession. If readers has a reasonably
background of Massive MIMO, e.g. antennas and wave good understanding of an area from start, they do not need to
propagation, the implications of Massive MIMO, e.g. architecture read everything in these chapters, and rather selectively read
and implementation and radio requirements are also covered. what is important to them.
Table of Contents
1. Antennas and wave propagation 4
2. Antenna arrays 12
3. Multi-antenna technologies 24
4. 3GPP—physical layer support for Massive MIMO 44
5. 3GPP—radio requirements 62
6. Architecture and implementation 74
7. Massive MIMO features 90
8. mmWave technology 124
9. Network performance 136

1. Antennas and wave

propagation
Ericsson | The Massive MIMO Handbook — Technology Primer 1. Antennas and wave propagation 5
Antennas and wave propagation

—introduction
The basis of wireless communication is the transmission and Wireless communication utilizes electromagnetic waves that
reception of radio waves using antennas. Traditionally, the way are polarized and can be superimposed on each other.
antennas have directed their transmission or reception has been
fixed and decided during manufacture or installation. However, Antennas are directive and can focus transmissions of waves
the introduction of antenna arrays and Massive MIMO allows the in specific directions. The larger the antenna is with respect to
transmission and reception capabilities to be varied dynamically the wavelength, the more focus and hence the higher antenna
in response to the spatial distribution of user and traffic and gain it can achieve.
the multiple ways that the radio waves propagate between the
Waves traveling from a transmitter to a receiver experience
transmitters and receivers.
path loss that tends to increase with increasing distance and
The properties of antennas and wave propagation are, therefore, frequency. This drives the need for more directive antennas at
fundamental to the design and operation of massive MIMO higher frequencies, preferably realized using Massive MIMO.
systems, which attempt to exploit these properties and overcome
Multipath propagation and the associated delay and angular
some of the challenges to reliable communication that they
spread affect numerous aspects of wireless communication,
present. This chapter introduces some key concepts of antenna
including Massive MIMO algorithm and hardware design and
theory and wave propagation, such as:
performance.
These concepts will be used in chapters [TP2-4] and [TP7-9].
1. Antennas and wave propagation

Properties of electromagnetic waves

No chang
Electromagnetic waves come in a large range 5G spectrum Electromagnetic waves

of frequencies and wavelengths 3G/4G spectrum are polarized
106 109 1012 1015 1018 Frequency [Hz]
Radio waves Infrared Ultraviolet X-rays and

gamma rays
Visible light
Many waves can be added together, ”superimposed” Superposition creates

standing wave patterns,
”fading”
+
=
+
The basis for wireless communication is the existence and A very important characteristic of electromagnetic waves is
use of electromagnetic fields. More specifically, by creating that they can be added together, ”superimposed”. The field
waves in these fields that travel through space, it is possible to vectors in a point in space are the sum of the field vectors from
transmit energy and information from one location to another each individual wave. Superimposing two or more waves with
location. The information is encoded into the amplitude, the same propagation direction can therefore create a new
phase, and frequency of the waves. The range of frequencies polarization. Any arbitrary polarization can be described as the
and wavelengths, the spectrum, of naturally occurring or superposition of two waves with orthogonal polarizations.
artificially produced electromagnetic waves span many orders
of magnitude. Mobile wireless communication systems up to Superposition of waves with different propagation directions
4G typically use waves with frequencies from a few hundred creates standing wave patterns where the average field strength
megahertz (MHz) up to several gigahertz (GHz), while 5G can is lower or even zero in some locations. This is referred to as
also utilize frequencies fading and has traditionally been a challenge to the reliability
of several tens of GHz. of wireless communications. However, the increasing use of
directional antennas that can distinguish the waves based on
An electromagnetic (EM) wave is transversal, meaning that their propagation directions is turning this into an opportunity
the electric field and the magnetic field are orthogonal to the instead.
direction of propagation and to each other. The polarization of
the wave describes how the electric field is oriented. A wave can
be linearly polarized if the field at a point in space is oscillating in
a fixed direction, circularly polarized if it is rotating with constant
amplitude, and elliptically polarized if it is both rotating and
oscillating.

Antennas
No
Antennas act as the interfaces

between conducted and radiated signals
... 1 0 1 ...
... 1 0 1 ...
The properties of the antenna (or array of antennas) wavelength wavelength

determine how waves are radiated or received in
different directions
Low directivity antenna Highly directive antenna
In wireless communication systems, antennas are used to Antennas that are small in relation to the wavelength usually
radiate electromagnetic waves from one location and capture have low directivity and therefore spread their radiated energy
some fraction of them in another location. Once the wave has in many directions. Antennas that are large in relation to the
been generated it will continue to propagate outwards, radiate, wavelength have more freedom in shaping the radiation pattern,
into the surrounding space. When the radiated wave encounters such as creating high directivity by focusing the transmitted
a receiving antenna the time-varying fields will drive a voltage or energy in a narrow range of directions. High directivity is very
current on the antenna terminals. Hence, the receiving antenna useful for improving the communication quality but also presents
captures some of the power that was transmitted from the other a challenge since the transmission and reception directions need
antenna. If the captured waves have enough power then the to be carefully aligned. Massive MIMO has evolved to solve this
receiver is able to decode the information carried by the waves. challenge in a dynamic environment with multiple moving users
and complex wave propagation.
The effectiveness with which antennas can radiate and capture
waves is fundamental to the wireless communication quality. Antennas of a given physical size tend to become more directive
All antennas are directive to some extent, meaning that they are and have more degrees of freedom for shaping the radiation
more effective in radiating waves in certain directions and with pattern as the frequency increases. Alternatively, for a given
certain polarizations, as illustrated by the radiation pattern. The directivity the antenna becomes physically smaller at higher
effectiveness or antenna gain is often given relative to that of frequencies. A consequence of this is that the ability of an
an ideal isotropic antenna that radiates equally in all directions. antenna to capture energy from a passing wave, its effective
Antennas are usually reciprocal, meaning that an antenna has antenna area, reduces with increasing frequency, assuming the
the same radiation pattern when receiving as when transmitting. same antenna gain at all frequencies. Therefore the use of more
Pairs of antennas with orthogonal polarizations are commonly directive antennas at higher frequencies becomes necessary
used in wireless communications to be able to transmit or receive which is another reason for introducing Massive MIMO.
waves with arbitrary polarizations.

Path loss
No chan
Only a tiny fraction of the transmitted power reaches Antenna directivity can help increase this power
the receive antenna Massive MIMO ensures that the transmissions occur in
the best directions
When an antenna radiates electromagnetic waves with a certain Directive antennas are regularly used to partially compensate
power into free space, these waves will expand spherically. As for the path loss and thereby increase the communication range.
the distance to the transmitter increases the power gets more This becomes particularly important at higher frequencies
spread out and hence the local field strength decreases inversely where the reduced effective area of the receive antenna and
proportional to the distance. The path loss (basic transmission the frequency-dependent losses in non-line of sight need to be
loss) is the ratio between the power transmitted by one isotropic mitigated. Massive MIMO ensures that the high directivity is
antenna and that received by another isotropic antenna. In free efficiently utilized by steering the radiation in the most beneficial
space propagation, the path loss is proportional to the square directions.
of the distance. The free space path loss is also proportional to
the square of the frequency, although this is not a ”propagation
loss” as such but rather a consequence of the reduced effective
antenna area of the receiving isotropic antenna.
When there are obstacles such as terrain or buildings that may

scatter or block some of the electromagnetic waves, the path loss
can increase quicker with distance, typically by the distance to
the fourth power for outdoor non-line of sight communications.
Many materials strongly attenuate radio waves passing through
them. Therefore, outdoor to indoor propagation losses can be
very high and also increase with frequency.

What is different at mmWave?

What is different at mmWave? I
mmWave
5G spectrum
Mid-band
1 50 100 Frequency [GHz]
3G/4G spectrum
Wider bandwidths available at high bands
Higher
propagation
losses
Can fit more antennas into array
without increasing the physical size
More spectrum is available in mmWave bands than at low/ When the radio waves diffract, i.e. bend around obstacles,
mid-bands, hence these bands are attractive for use for they lose slightly more power at higher frequencies compared
communication. The wider bandwidths can be used to support to lower frequencies. Some materials strongly affect the
higher peak rates or to increase the capacity by allowing more transmission of radio waves leading to much higher losses
simultaneous users to access the radio network. at higher frequencies. Others, such as clear glass, are quite
transparent at any radio frequency. Similarly, reflections and
Since the wavelength becomes shorter with higher frequency, diffuse scattering tend to be of similar strength over a wider
the size of a resonant antenna (such as for example a half- frequency range.
wavelength dipole) also reduces correspondingly. Therefore,
it becomes possible to fit more antenna elements in the same These factors mean that achieving full coverage is more
physical antenna array size. This can give higher directivity challenging at higher frequencies, especially in deep non-line
which will allow more power to be transmitted or received in of sight or outdoor to indoor. Realizing the full potential of the
a particular direction. At the same time, the beams become beamforming is necessary to approach similar coverage as for
narrower and require more precise beam steering. This also lower frequency bands.
presents challenges for UE design where coverage in all
directions is needed.
As mentioned previously, the effective antenna area, i.e. the

ability of the antenna to receive power from an impinging
electromagnetic field, likewise scales with the wavelength.
Unless balanced with maintaining the physical size of the
antenna array, this will lead to a reduction in received power
at higher frequencies.

Multipath propagation creates

fast fading No chan
Massive MIMO mitigates and exploits fast fading

Limited angular
spread
Wave-
length Large angular Wave-
spread length
As the electromagnetic waves travel along multiple propagation A moving transmitter or receiver will experience rapid time
paths from the transmitter to the receiver we speak about variations of the signal strength due to the fading, often
multipath propagation. The superposition of many waves give changing completely within a fraction of a second. This kind of
rise to fading, i.e. variations in time, space, and frequency of the fading is often referred to as fast fading, and has always been
average field strength and polarization. The larger the range a challenge to reliable and efficient communication. Modern
of directions in which propagation paths occur, the angular wireless communication uses opportunistic scheduling and link
spread, the more rapid the variations are. A typical user often adaptation to ensure that the fading channel is utilized to the
experiences multipath from many directions causing the greatest extent possible.
distance between peaks and nulls in the fading pattern to be
less than half of a wavelength. At a base station which is often The fading variations are not only confined to time and space.
elevated above its surroundings to provide large area coverage, The different path lengths of the different multipath will cause a
the angular spread is typically confined to a more narrow interval delay spread of any transmitted signal. The different delays will
and therefore the peaks and nulls gets spread out more. cause the superposition of the waves to change if the frequency
is adjusted slightly, leading to fading variations also in within the
The multipath and angular spread strongly impacts how antenna frequency band in which communication is occurring.
directivity can be used and whether simple beam shapes are
adequate or more irregular radiation patterns are needed to Fast fading in space, time and frequency has always been one
optimize the communication. Massive MIMO contains methods of the main challenges to maintaining reliable communication
and algorithms for acquiring information about complex and quality, particularly for high bitrates with low latency. Multi-
rapidly changing multipath conditions and using the degrees antenna techniques started out as methods to mitigate the
of freedom of electrically large antennas with dual-polarized fading by adding diversity but has subsequently evolved to more
elements to phase shift different multipath components such directly take advantage of the multipath channel by potentially
that they add constructively and enhance the communication using the different propagation paths and polarizations as
quality. separate and parallel communication channels through MIMO
schemes.


2. Antenna arrays
2. Antenna arrays
Ericsson | The Massive MIMO Handbook — Technology Primer 2. Antenna arrays 13
Introduction to antenna arrays

Introduction to antenna arrays Com
Not
Antenna array Subarray Dual-polarized

element pair
Massive MIMO generally uses planar arrays of dual polarized

element pairs divided into subarrays
In this chapter, a basic description of antenna arrays and what A subarray is the smallest dynamically controllable entity of the
can be accomplished with them are presented. In the figure antenna, and each subarray has two associated radio chains,
above, an example with a 128-element uniform planar array one per polarization. By partitioning the array into subarrays,
with eight rows and eight columns of dual polarized element as is also illustrated in the figure above, it is possible to reduce
pairs is given. Such as uniform and planar structure is the most the number of radio chains. The chapter also gives some
common structure in Massive MIMO deployments, and the background for how subarrays are often used in practice.
description in this chapter is limited to this array type.
Finally, antenna arrays can be used to adapt the beam shapes
The basic principle for beamforming with an antenna array and for spatial multiplexing, denoted MIMO (multiple-input
and a beamforming weight vector, which can be used to form multiple-output). There is spatial multiplexing both to a single
a beam, are introduced. More specifically, so-called classical user, SU-MIMO, and to multiple users, MU-MIMO.
beamforming is taken as a baseline in this chapter.
Terms such as main lobe and side lobes, which are often used
when describing radiation patterns, or beam patterns, are
defined, and it is explained that the patterns can be expressed
as the product of an element factor and an array factor. Some
properties of the beams as a function of for example the number
of array elements are then stated.
2. Antenna arrays
The basic beamforming principle is to

transmit the same signal from multiple No cha
antennas
Zero wavelength
+ = difference gives
maximum gain
Partial wavelength
lower gain
Half wavelength
cancelation (null)
The electromagnetic field generated by an array is a super- If all the signals have an identical time delay, or phase, on the
position, i.e. a sum, of the contributions from the individual receiver side, completely constructive addition is achieved and
elements, [TP1, p. 7]. This means that a beam may be formed maximum gain in signal strength is obtained. In the example
by transmitting the same signal from multiple antennas. It takes above, this happens right in front of the array (top). Conversely,
time for the signals to propagate to the receiver which thus sees a if the signals are completely out of phase, they cancel in the
delayed signal. The signals from the antennas add up over the air summation and a zero signal is obtained which corresponds
and appear as a sum signal on the receiver side. This is illustrated to zero gain (bottom). For directions in-between these two
above assuming a free-space channel and the simplest case with extremes, the phase difference is somewhere in-between and
two antennas above. so is the resulting beamforming gain (middle). At such angles,
the maximum gain in signal strength is between zero and the
maximum gain, which in this case is two.
The far-field superimposed signal (in free-space) as a function

of direction (spatial angles) is often referred to as a beam
pattern. The beam pattern is generally a complex-valued
function that contains both amplitude and phase information.
For classical beamforming it is common to look at the amplitude,
which will display the spatial beam associated with the radiation
pattern.
2. Antenna arrays
Classical beamforming directs power

in desired directions by adjusting relative
Classical beamforming directs power in desired
delays
directions by adjusting relative delays
de de
litu litu
Copies of the signal amp amp
1X 4X
with different delays
+=
𝜃𝜃0
𝑤𝑤1
𝜃𝜃0
signal 𝑤𝑤2
𝜃𝜃0
In direction 𝜃𝜃0 "ar away from
𝑤𝑤3
the array the element signals appear
𝜃𝜃0 as plane waves in phase
𝑤𝑤4
Weights controlling the relative delays between elements to align in direction 𝜃𝜃0
By transmitting copies of the same signal and appropriately For classical beamforming, the beamforming weight vector is
adjusting the different amplitudes and delays from all the chosen to maximize the transmitted signal power in a desired
elements it is possible to control the radiation pattern. In fact, direction. Such a choice is optimal in free-space single-path
since the total delay for the signal from each antenna is the sum propagation with LOS between to the receiver. Classical
of the adjusted delay and the propagation delay it is possible beamforming is illustrated above for transmission of a single
to control the direction with maximum gain stemming from frequency sinusoid in a direction θ0. where the contributions from
constructive addition of the signals from all antennas. In other all the elements appear as plane waves in phase. Again,
directions, the contributions add destructively and when they the superposition is a sinusoid with the same frequency but with
cancel out there will be nulls. an amplitude which is the sum of the amplitudes i.e., four times
the amplitude as compared to a single element for an array with
For 4G/5G massive MIMO systems, the delays needed are four elements.
small, and a delay of a signal is equivalent to a phase shift of
the signal. The amplitude and the delay adjustment of a signal Above, transmission was described. The array may also, just as
may therefore be represented by a complex number, or weight. any antenna in general, be used for reception, [TP1, p. 7], and it
The set of weights for all the antennas is often collected in a is then possible to amplify desired signals received from certain
beamforming weight vector, where each element of this vector directions and suppress, or null, interfering signals in other
represents the delay and amplitude of that specific element. In directions.
the figure above w1, w2, w3 and w4 are the element adjustments
and the beamforming weight vector w would be a vector defined
from these elements.
2. Antenna arrays
Beam patterns typically have multiple

lobes
No chang
3D spherical plot Vertical cut (y=0)
Main lobe
Side lobes
Gain (dBi)
Vertical angle
A radiation pattern describes properties of the electromagnetic For classical beamforming, introduced on the previous slide,
field as a function of direction, and a beam is often described the beam looks like a ‘beam’ in the intuitive sense of a ‘directed
by its radiation pattern. Most often, the power of the field is entity’. By this we mean that the beam pattern has a single
illustrated in the form of gain relative an isotropic radiator, and well-defined main lobe with maximum gain and sidelobes with
this referred to herein as a beam pattern. significantly lower gain levels, similar to a flashlight beam.
To the left above, a 3D spherical plot of such a gain pattern is As a sidenote, the term “beam” of antenna was originally defined
exemplified. Often, cuts of such a pattern are also illustrated and as the main lobe of the beam pattern of an antenna. With
to the right, a vertical cut in the xz-plane y=0 of the same pattern generalized beamforming in multipath propagation channels,
as function of the angle θ is shown. In here, θ is the angle to the however, not only the main lobe but the entire beam pattern is of
positive z-axis, also referred to as vertical angle, and the beam interest. This will be further elaborated in [TP3, p. 27]. The reader
pattern illustrates a beam steered to direction θ0=90°, i.e., the is referred to [1, Ch. 5] for a more thorough discussion.
direction forward, perpendicular to the antenna array.
Beam patterns typically have multiple lobes. Lobes are different

regions of the beam pattern separated by nulls, and a null is a
direction in which the contributions from the different antennas
add destructively. A main lobe is a lobe in which the gain attains
its maximum. Other lobes, where the gain does not attain it
maximum, are referred to as side lobes.
2. Antenna arrays
Beam patterns depend on the number

of elments and the element pattern No
With N elements, the maximum gain over an element is N and the main lobe width scales as 1/N
Gain (dBi)
Gain (dBi)
Gain (dBi)
Gain (dBi)
Vertical angle Vertical angle Vertical angle Vertical angle
The element gain pattern shapes the envelope over the set of beams
Gain (dBi)
Gain (dBi)
Gain (dBi)
Gain (dBi)
Vertical angle Vertical angle Vertical angle Vertical angle
For an array, the total electrical field is the sum of the For classical beamforming with an N-element uniform linear
contributions from the individual elements. This means it array, the maximum gain is N and the width of the main lobe
depends not only on the relative amplitude and delays between is proportional to the inverse of the number of elements, ~1/N.
the different copies from the individual antennas but also on the This means that as the size of the array grows in terms of the
radiation patterns of the elements. On a logarithmic scale, the number of elements, the maximum gain increases but the main
total gain is the sum of the elements’ gain and the array gain, lobe becomes narrower. This is illustrated in the top figure
above, and the reader is referred to [1, Ch. 4] for more details and
Total gain pattern [dBi] = element gain pattern [dBi] + array gain derivations.
pattern [dB].
Since the main lobe becomes narrower with increasing number
Gain patterns rather than just gain are used to highlight that of antennas, dynamic beamforming or UE specific beamforming,
the equation holds in all directions and not only in a single as discussed in [TP3, p. 26] becomes necessary. This mean that
direction. It is furthermore assumed that all the elements of the the direction of the main lobe can be pointed to where the UE
array have identical radiation patterns so that there is a single is when it is scheduled. Furthermore, since the direction of the
element pattern. The array gain pattern on the other hand main lobe in which the array gain pattern has its maximum N
captures how the contributions combine in different directions can be steered by changing the beamforming weight vector, it
for a given beamforming weight vector. In directions where the follows from the relation above that the element gain pattern
contributions add constructively, the array gain is high and in shapes the envelope over the set of beams, or maximum gain,
directions with nulls the array gain is zero (on linear scale). that can be generated.
2. Antenna arrays
Vertical or horizontal linear arrays

—same maximum gain but different
No chang
beam patterns
Vertical arrays
The higher the array, the narrower
the beam in the vertical dimension
Horizontal arrays
The wider the array, the narrower
the beam in the horizontal plane
It is possible to form both vertical arrays as well as horizontal compared to a single element depends on the number of
arrays by stacking elements on top of each other and next to each antennas. In the illustrations above, the maximum gains are
other respectively. obtained along the x-axis and are the same for the same number
of elements for both the vertical and horizontal arrays. Their
Assuming again classical beamforming and uniform linear arrays, shapes are however different.
the width of the main lobe will change in both cases:
Note that conventional sector antennas are implemented as
• For a vertical array, the vertical main lobe width, which is the vertical arrays, for example as a single column of dual-polarized
width of the main lobe in the vertical cut (y=0 above) scales elements. Such a single-column antenna could have two
with the height of the array. The taller the array, the narrower connectors, one for each polarization, feeding all elements with
the main lobe in the vertical dimension. the same polarization through a feeder network. Since the feeder
network is not dynamically changed, one can think of this as
• For a horizontal array, the horizontal main lobe width (in the
static beamforming and the maximum gain of the sector antenna
cut z=0 above) scales with the width of the array. The wider the
as compared to an individual element is then given by the number
array, the narrower the beam becomes in the horizontal plane.
of elements in the column. So, a tall antenna with many elements
Recall from the previous slide that the maximum gain as has a high gain, but at the same time, the width of the main lobe
in a vertical cut is narrow. This follows from the previous slide.
2. Antenna arrays
Planar arrays enjoy the benefits of both

vertical and horizontal arrays
No
+9dB
+9dB
Gain (dBi)
Vertical angle Horizontal angle
More antenna elements lead to narrower main

lobe in both vertical and horizontal dimensions
A linear array can be either vertical or horizontal, and by 10 log10Nv Nh=10 log10Nv +10 log10Nh
increasing the sizes of such arrays, the main lobe width in the
vertical or horizontal dimension will decrease while the gain which can be recognized as the sum of the gains for the
grows. A uniform planar array consists of elements arranged in corresponding vertical and horizontal arrays. Thus, classical
a uniform grid with Nv rows and Nh columns. As mentioned in beamforming with an 8x8 planar array offers an array gain
the introduction, uniform planar arrays are the most common of 18 dB which can be thought of, for example, as 9 dB from
configurations used in practice for Massive MIMO. beamforming with an 8x1 vertical array of elements which all
enjoy a 9 dB array gain using 1x8 horizontal arrays.
Two ways to view such an array include
Furthermore, the width of the horizontal main lobe scales with
• A vertical array where each element is a horizontal array ~1/Nh, whereas the vertical scales with ~1/Nv. Thus, in the
case with a planar array the main lobe widths decrease in both
• A horizontal array where each element is a vertical array dimensions when the number of elements increases in both
Recall that the total gain pattern (on a logarithmic scale) is the dimensions.
sum of the element gain pattern and the array gain pattern.
For classical beamforming, the maximum of the array gain
pattern will be given by the total number of elements,
2. Antenna arrays
Partition the array into an array of subarrays

to reduce the number of radio chains No change
Vertical cut Vertical cut of

For an antenna of fixed size, the smaller
of subarray beam envelope over array beams
the angular range requiring high gain,
the fewer radio chains required 32 subarrays 6 dB
→ 4a
4a
2x1 subarray 64 radio chains
Antenna array
4x1 subarray 3 dB
16 subarrays
2a 2a
→
32 radio chains
8x1 subarray
Dual-polarized 6 dB
8 subarrays a
element pair a
→
16 radio chains
Uniform planar arrays of dual polarized antenna pairs are To see this, first note that the maximum gain is proportional
commonly used. The two polarizations can be used for spatial to the antenna size in terms of the number of elements. So,
multiplexing, for diversity and to mitigate losses due to partitioning the array into subarrays does not reduce the
polarization mismatch. maximum gain achievable.
A uniform planar array can be partitioned into an array of However, the larger the subarrays, the narrower its half power
subarrays (AOSA). A key benefit of such a partitioning is that the beamwidth and since the total gain pattern is given as the
number radio chains needed is reduced compared with one radio product of the array gain pattern and the subarray pattern
chain for every element). Instead of one pair of radio chains for (which is the element in the array of subarrays) it follows that
each dual-polarized element pair, only one pair of radio chains the envelope of the beams, which describes the achievable total
is needed per subarray. The disadvantage of the grouping of gain, also becomes narrower.
elements into subarrays is that the range of angles within which
the beamforming can be done without significant gain drop In the figure to the right above, partitioning of an eight-element
becomes narrower. array into subarrays of different sizes is illustrated, two elements
at the top, four elements in the middle and eight elements at the
bottom.
More reading can be found in Advanced Antenna Systems for 5G

Network Deployments, [1], and Advanced Antenna Systems for
5G networks, [2].
2. Antenna arrays
Partition the array into an array of subarrays

to reduce the number of radio chains
Short ISD
High rise buildings
Mid ISD
Low/Mid rise buildings
Large ISD
Flat buildings
In practice it is of interest to have as large subarrays as possible

while ensuring that most of the UEs are within the main lobe of
the subarray. This offers high gain while not requiring too many
radio chains. In practice, vertical subarrays are used, and the
choice of subarray size depends on the UE and channel angular
distribution.
• In dense urban high-rise scenario, a small subarray such as 2x1

is preferable. This is so since the angular spread [TP1, p. 10] of
users in the vertical domain is large.
• In suburban rural scenario with more sparse deployment of

base stations, larger subarrays such as 8x1 can be used due to
the small angular spread of users in vertical domain.
2. Antenna arrays
Antenna arrays can be used for spatial

multiplexing, both single-user MIMO and
multi-user MIMO
Antenna arrays can also be used to transmit multiple streams, or Suppose that two signals s1 and s2 are to be transmitted and that
layers, simultaneously with different beams on the same time- each of them has a beamforming weight vector to generate the
frequency resource. This is referred to as spatial multiplexing and corresponding beam. The signals transmitted from each antenna
the streams of data symbols multiplexed are referred to as layers. is then taken as the sum of contributions for the two signals
and the signals transmitted from all the antennas can then be
There are two basic use cases expressed as
• Single-user MIMO, (SU-MIMO) where multiple layers are s=w1s1 +w2ss
transmitted to a single user terminal. This requires a multipath
propagation channel as well as a receiver with multiple receiver Here w1 and w2 are the beamforming weights for the two signals
antennas. and the vector s represent the signals transmitted from all the
antennas. Put another way, to transmit multiple signals, a sum
• Multi-user MIMO (MU-MIMO) where multiple layers are of signals is transmitted from each antenna rather than a single
transmitted to different user terminals in different directions. signal. This most often means that contributions from all signals
are transmitted from all the antennas.
Spatial multiplexing is further discussed in [TP3, p. 32-41].
2. Antenna arrays
Ericsson | The Massive MIMO Handbook — Technology Primer New image 0012. Antenna arrays 23
Summary
• For classical beamforming, more antenna elements

results in a narrower main lobe with higher gain
• An array can be partitioned into an array of sub-
arrays, with short vertical subarrays in small cells
and tall subarrays in large cell
• An array can be used to transmit multiple
streams of data at the same time
This chapter introduced concepts such as main lobe and side An array can be partitioned into subarrays to reduce the number
lobe which are used when discussing radiation patterns which of needed radio chains since high beamforming gain is most
are referred to as beams. often only needed in a limited angular range.
For classical beamforming using a planar array with N and M • The larger the subarray, the narrower is the range of angles
elements in the vertical and the horizontal dimensions, with high gain,
respectively:
• In practice, vertical subarrays are used, and in large cells,
• The maximum gain in the main lobe as compared to a single larger subarrays can be used and in small cells small subarrays
element equals the number of elements in the array, that is NM can be used.
• The main lobe has a width proportional to 1/N and 1/M in the An array can be used to transmit several streams with different
vertical and horizontal dimensions, respectively. beams at the same time.
• Typically, all streams are transmitted from all antennas
• Both to a single user as well as to multiple users (SU-MIMO

and MU-MIMO)
2. Antenna arrays
3. Multi-antenna
technologies
3. Multi-antenna technologies
Ericsson | The Massive MIMO Handbook — Technology Primer 3. Multi-antenna technologies 25
Introduction
Static versus UE-specific Beam-pattern based analysis Nullforming and beam pattern
beamforming and pitfalls
Low gain
Low gain Intentional

null
Spatial multiplexing for SU- Channel info needed for beamforming MU-MIMO – realistic expectations?
and MU-MIMO
Performance:
1
Layer 1 Layer
Layer 2 Layer 2
Layer 3 Layer 3
This chapter deals with miscellaneous topics concerning Spatial multiplexing for SU- & MU-MIMO: Spatial multiplexing
Massive MIMO technologies with the intent to give a more can be achieved in several ways depending on the antenna setup
detailed understanding of selected topics than the more and UE locations. Understanding how it works is important for
high-level exposition in the main story. making appropriate choices of technologies and deployment.
Static versus UE-specific beamforming: A key feature of a Channel info needed for beamforming: Information about
Massive MIMO system is its ability to tailor the beam pattern for the channel between the base station and the UE is needed
each UE instead of as in classic systems where many UEs share to know where to beamform. Collecting such information can
the same beam. The difference between such UE-specific and be done in various ways, largely categorized as codebook or
static beamforming needs to be understood. reciprocity based. The two categories have different strengths
and weaknesses.
Beam-pattern-based analysis and pitfalls: Beam patterns
are commonly used to describe transmission and reception MU-MIMO—realistic expectations?: Marketing of MU-MIMO
techniques in Massive MIMO systems. It is a simple and intuitive with many layers has at times, been very strong. The use of demo
tool but a source of much confusion if its limits are not fully setups in the industry, as well as analysis in the literature, easily
understood. give the impression that MU-MIMO is a ubiquitous technique
with extraordinary benefits. Get to understand why industry
Nullforming and beam-pattern: Nullforming is a key demos and the literature tend to overestimate the benefits due
beamforming technique especially important for MU-MIMO to assumptions which are rather far from realistic commercial
to suppress interference towards victim co-scheduled Ues. network deployments. Setting realistic expectations and
The properties of the multi-path environment affect where understanding in what subset of cells MU-MIMO makes a real
co-scheduled Ues can be located relative to each other for impact is an important takeaway.
efficient MU-MIMO operation. The quite intricate interaction
between beam pattern, multi-path, nullforming and MU-MIMO,
therefore, deserves a more detailed treatment.
Static vs. UE-specific beamforming
(Semi) Static beamforming UE-specific beamforming

• The same fixed beam is shared by many users • Each UE gets a beam of its own
• All UEs in cell cell- specific beamforming • No sharing so beam can be tailored for each UE
• Users often miss the beam peak – gain suffers • Beamforming gain can be maximized
• Example: A fixed sector beam with some down - tilt
Low gain
Low gain
The beam pattern can be static or vary more or less dynamically. The horizontal spatial UE distribution is however typically
Beamforming that dynamically changes beam pattern in entirely different with often a wider and rather uniform UE
time and frequency is typically exploited to provide so-called angular spread over the sector. This means that in horizontal
UE-specific beams that are tailored for each UE. In contrast, direction beamforming gain is lost with a static beam approach,
static beamforming has a beam pattern that is constant over either since the static beam must be made wide to cover all UEs
time and is a reasonable simplification when all UEs in a cell or because a narrow beam misses most of the users.
can share the same beam. Semi-static beamforming allows the
beam pattern to change over time, but substantially more slowly UE-specific beamforming
than dynamic beamforming so that it essentially is constant over A key benefit of Massive MIMO systems is their ability to form
long periods of time. Automatic adaptation of beam pattern for high gain UE-specific beams towards UEs of interest. Thus,
cell shaping (c.f. [TP7.3, p. 114]) is an example of the latter. there is possibility to tailor the beam pattern to suit a particular
UE at each moment in time. This allows the beam to be narrow
Static beamforming and still track movement of the intended UE. Base station side
The use of a static beam pattern in elevation is common in beamforming allows each UE in the cell to enjoy a specific beam
cellular networks. For example, a single-column antenna pattern with a strong peak instead of being forced to share the
whose beam pattern is fixed and with a narrow main lobe in same beam over many geographically dispersed UEs as in the
the vertical domain and wide sector covering in the horizontal static beamforming case. This leads to higher SINR levels and
domain is often used to serve all the UEs in a cell. To obtain good thus better overall system performance.
beamforming gains with a fixed beam system it is important that
all the UEs that are intended to be served by the fixed beam are
located in a direction which is close to a strong peak of the beam,
otherwise beamforming gain quickly drops. With above-rooftop
deployments and large cells this is often approximately fulfilled
in the vertical domain for a relatively large portion of the UEs in
a cell; as seen from the base station many UEs tend to be close to
the horizon so the vertical UE angular spread is relatively small.
What is a beam? Two beamforming types

No
Classical beamforming Generalized beamforming
• Does not look like a “beam” – may have many lobes

• Well-defined main lobe – single pointing direction • Beamforms into all the propagation directions
• Good for low angular spread – optimal for free-space of the channel
• Optimal for multi-path channels with angular spread
Beamforming may be classified into the two categories multi-path channels, leading to so-called frequency-selective
of classical beamforming and generalized beamforming. beamforming. There can be many lobes and it can be hard to
Such a classification is useful to help resolve common identify a single dominating main lobe. A somewhat simplified
misunderstandings concerning radiation/beam pattern and explanation for this is that the beamformer now needs to emit
what constitutes power in all of the plethora of propagations paths and hence the
a “beam”. beam pattern resembles the distribution of “channel directions”.
To be more precise is difficult, since the relative phases of the
Classical beamforming radio waves on the various propagation paths and how they
Classical beamforming is easy to understand and agrees well combine now start to matter. As a consequence, analysis via
with intuition. The beam pattern has a well-defined and often beam pattern provides limited understanding on how the
a relatively narrow dominating main lobe which can be pointed communication link will perform.
into a desired physical direction, for example from a base station
towards a UE. Classical beamforming is the focus of traditional With significant channel angular spread, classical beamforming
antenna theory, which is covered in more detail in [TP2]. In is no longer optimal and leads at least in theory with ideal
free-space or line-of-sight propagation, classical beamforming channel knowledge to an SNR loss as it is unable to match the
is, or is near, optimal in the sense of maximizing signal strength fast-fading properties of the channel.
achieving maximum gain. Such channels have (essentially) a
single propagation path, i.e., zero channel angular spread, and It is surprisingly difficult to give a good, relevant, and precise
no or little multi-path fading as discussed in [TP1, p. 10)]. Thus, definition of what a beam is. An attempt to nevertheless do so
a beam with a single direction matches the channel very well. is found in Section 5.5 of [1]. It suffices to say here, and roughly
speaking, that a beam is used for transmitting/receiving a
Generalized beamforming stream of symbols, such as a layer. So a layer is transmitted/
In contrast, generalized beamforming includes totally arbitrary received on a beam of it own. Note in particular that the shape
beam patterns with potentially no obvious connection to of the beam, i.e., the characteristics of its radiation pattern, as is
propagation path directions in the channel. In contrast, also evident from above not part of the definition, although the
generalized beamforming includes totally arbitrary beam particular type of beams illustrated in this handbook generally
patterns with potentially no obvious connection to propagation have a corresponding radiation pattern.
path directions in the channel. These beam patterns typically
vary also over frequency to track the frequency-variations of
Optimal beam patterns may have weak

correspondence with propagation directions
No chang
Propagation directions: 0, 10, 40 degrees Propagation directions: 0, 10, 40 degrees Propagation directions: uniform 360 degrees
1:st set of random propagation phases 2:nd set of random propagation phases 3:rd set of random propagation phases
Ok correspondence with beam peaks Bad correspondence with beam peaks No correspondence with beam peaks at all
Phases matter but not visible in beam patterns!
It is reasonable to think that a beam optimized for a certain However, consider now the middle figure, where again the same
channel would clearly show the propagation directions as three propagation directions as in the left figure are used. Despite
corresponding peaks in its beam pattern, in-line with classical identical propagation directions, the beam pattern looks vastly
antenna theory as in [TP2, p. 15-19)]. This is however far from different and no longer at all matches for two of the propagation
always the case as will be demonstrated next. directions where only a single lobe peak is present instead of two
distinctly different peaks as in the left figure. This qualitatively
Consider a multi-path scenario with three distinct propagation substantial difference in optimal beam pattern is only due to the
directions and the use of generalized beamforming. In the left difference in the relative phases of the three propagation paths
figure, the three lobes of the beam are seen to point roughly in between the two figures. The relative phases between paths
those three propagation directions. This confirms the intuition determine whether signals of those paths combine constructively
that an optimal beam pattern should match the distribution of or destructively, which may have a major impact on the beam
channel directions. Clearly, the use of classical beamforming in pattern. But the phase characteristic is not observable in a beam
this case would not be optimal and lead to a lower SNR as it is pattern analysis, which greatly diminishes the value of such an
unable to match multiple propagation directions at once. analysis in cases of significant channel angular spread.
Unlike in a single-propagation path case, the lobe peaks do not This analysis method completely breaks down in the extreme
exactly match the propagation directions. This is because the case of 360 degree uniformly distributed propagation paths as
array has a finite resolution so is unable to completely resolve shown in the right figure. The beam pattern then has absolutely
individual channel directions and in particular is unable to nothing to do with propagation directions of the channel.
resolve directions that are very close to each other. Nevertheless,
there is a fairly good and easy to understand correspondence
between channel directions and beam shape.
Nullforming and relation to beam pattern.

Raises SINR by lowering interference part
N
Z
W
Classical nullforming Generalized nullforming
• Angular spread is small • Angular spread not small, hence phases of paths matters
• A beam pattern null points towards a victim UE • Direction to a victim UE may not correspond to
a beam pattern null
Beam pattern completely informative! Beam pattern not informative!
Nullforming is a technique for on purpose reducing the Generalized nullforming

beamforming gain in certain “directions” so as to reduce Generalized nullforming is a potential component of generalized
interference at intended points in space. For example, reducing beamforming. Just like the latter, generalized nullforming
unwanted inter-cell interference stemming from transmissions targets and supports also the multi-path propagation scenario.
from a base station that hits a cell-edge UE in a neighboring As expected, the beam pattern using generalized nullforming
cell. Receive side nullforming is very common both in UEs may look rather arbitrary and non-intuitive. In fact, it may be
and in base stations while transmit side nullforming is lately even more chaotic than for generalized beamforming without
becoming increasingly popular in the base station primarily due nullforming. Since there are several propagation paths, lobes
to the introduction of MU-MIMO. Similarly, as for beamforming, in different parts of the beam pattern may form signals that,
nullforming can be categorized into a classical and a generalized by carefully crafted phase characteristic, add destructively at a
flavor. specific location in space. This relies on fast fading properties and
therefore a null at one location in space may turn into a strong
Classical nullforming signal at another location just some fraction of a wave-length
For classical nullforming, the physical directions of the nulls in away. Nullforming is therefore a technique which is sensitive to
the beam pattern each point towards some victim UE. This make errors in the channel state information (CSI), including due to
intuitive sense and corresponds to a scenario with no or very noise and doppler, and thus needs good coverage locations and
little channel angular spread. In such a scenario, nullforming low mobility.
also ensures there is a zero or low signal level at the location of
the UE. The left figure depicts the beam pattern of such classical More specific examples of classical and generalized nullforming
nullforming and it is seen how the red dashed lines, showing will be given on the two following pages.
directions of three different victim UEs each align perfectly with
a corresponding null in the beam pattern. This particular beam
is targeting a UE in a direction along the green dashed line. It is
seen that although the present scenario assumes a free-space
channel, the direction of the UE of interest is not along the
peak of the lobe, illustrating how nullforming may decrease the
beamforming gain for the served UE by offsetting the peak.
Nullforming for MU-MIMO,

—Free-space No chang
• Want strong signal for UE 1 and nulls for the others

• Free-space propagation
• Channel does not distort signals
• Nulls on UE side and on BS side now coincide

• Peaks as well UE1 UE2
UE3 UE4
• Beam-pattern based analysis totally
relevant
• Sufficiently low angular spread required
Low angular spread makes things intuitive!
The figure considers a free-space scenario where MU-MIMO is The interfering received signal contributions at the three victim
used to transmit four layers, each to one of four different UEs UEs is however zero since null forming is used to form a beam
1 – 4. The blue surface represents the received power level, pattern so that nothing of the transmitted signal intended for
where the height of the surface is proportional to the received UE 1 leaks to the victim UEs. Due to the free space assumption,
power level at the corresponding position in the horizontal there is a perfect correspondence between the beam pattern
plane. Although the received signal is a superposition of the four and the resulting received signals and it is seen how each
different transmitted layer signals, the figure only shows the victim UE direction has a null in the beam pattern. Although the
transmitted beam pattern and received signal that is intended beam diagram looks much more complicated then for classical
for UE 1, the UE marked in green. The transmitted signal spreads beamforming, it still makes intuitive sense.
all over the horizontal plane as shown by the blue surface,
potentially creating interference to the three other The intention of null forming is to create a strong signal for the
victim UEs 2 –4. served UE 1 with small leakage to the victim UEs. However, it is
here seen that UE 1 has a close to zero received signal. This is
due to that UE1 is very close to UE2, and the latter has by the null
forming design a zero received signal. It is in the present case
physically impossible to have a strong signal at the UE 1 location
when the requirement is to have a zero signal level just next to it
at the UE 2 location.
Nullforming for MU-MIMO

—Multi-path
Nullforming for MU-MIMO – Multi-path
• Nulls on UE side but not always on BS side

• Thinking in terms of beam pattern is a useful tool
• When its limitations are understood LOS directions of UEs
• Otherwise, a source of much confusion
UE1 UE2
• No multi-path: beam pattern is intuitive
UE3 UE4
• Multi-path: beam pattern non-intuitive
and arbitrary
Multi-path with MU-MIMO leads to real mess!
This figure illustrates exactly the same setup as in the previous Roughly speaking, since a propagation path reflection point is
page, but now in a scenario involving significant multi-path. typically very far away from the transmitting antenna array, from
Thanks to the nullforming, the three victim UEs 2– 4 still have the perspective of reception the reflection point acts as creating
zero received interfering signal. However, the beam pattern is a very large transmit antenna array and such a large array can
no longer a good predictor for what really matters – the received produce very narrow lobes (c.f [TP2, p. 17]) which avoid the
signal levels. Although UE 4 has indeed a null in the beam problem of a nearby null necessarily limiting the signal strength.
pattern, UE 2 and UE 3 instead have peaks! This may seem
confusing, but due to the multi-path several strong lobes on the Clearly, the beam pattern is in this multi-path case quite
transmit side may now convey the signals that after propagating misleading. The use of MU-MIMO involving nullforming tends
over the channel combine at the UE location to cancel each other. to result in messy beam patterns to an even larger extent than
The phase characteristic of the beam now plays a crucial role to what’s caused by the multi-path alone. Although a beam-pattern
achieve cancellation but is not observable from the diagram. based analysis is an often used and useful tool, its limits need to
be understood. Otherwise, it is a source for much confusion and
Although hard to discern in the figure, the served UE 1 has now in possible harm. As an example, there is a risk of over-emphasizing
fact a strong received signal, despite the proximity to the null at the need of requirements on various beam pattern properties,
the UE 2 location. The presence of multi-path again provides the with suboptimal designs for reality as a potential negative
explanation. Just as signals from different lobes may now follow consequence.
different propagation paths and still reach the same victim UE
and cancel each other, with proper phase characteristics signals
can now constructively combine at UE 1 location.
SU-MIMO spatial multiplexing
How does it work?

• Several layers to the same UE
• “One layer/beam per significant propagation path”
• Channel angular spread needed for more than 1
layer per polarization
•Dual - layer transmission in LOS on separate
polarizations important use case
• Needs at least as many transmit/receive antennas
as layers
• Fast fading based nullforming can reduce
inter - layer interference
Ubiquitous benefits for both coverage and capacity!
Spatial multiplexing for SU-MIMO is a way of increasing system The above mental model holds well if directions of the
capacity via improved link performance by multiplexing several propagation paths are very well-separated relative to the beam
layers of data streams on the same time/frequency resources. widths. This is however rarely the case and hence the beams
can overlap in arbitrary ways and the separation of the layers
To roughly understand the underlying technical principle, to a large degree instead hinges on fast fading properties of the
for simplicity, temporarily assume single-polarized antenna channel where phase relations are crucial but hard to understand
setups. A possible mental model of spatial multiplexing where via a physical direction-based analysis (c.f. the limitations
all the layers are transmitted to a single user is that each layer of beam pattern analysis for generalized beamforming/
is transmitted on a separate beam and the beam is matched nullforming). Generalized nullforming is then a key tool in
towards a particular and significantly strong propagation path. separating the layers so as to reduce inter-layer interference. On
This is illustrated in the top figure and it is easy to understand the receive side, such nullforming is basically always used while
that transmitted layers can be separated out on the receive side in more advanced cases generalized beamforming/nullforming
by means of receive beamforming (spatial filtering) if those is also used for transmission, e.g. in the downlink from the base
layer-carrying paths have different arrival angles. Thus, one station with reciprocity based TDD or with advanced codebook-
layer/beam per propagation path and you need multiple such based feedback [TP4, p. 52].
paths to be able to convey multiple layers – channel angular
spread needs to be sufficiently large. The maximum number of In practice, it is common to use dual-polarized antenna setups
layers is ultimately limited by the number of base station/UE at both the base station and UE sides, see the bottom figure
antennas, whichever is smallest. Note that the beams here, and which shows dual-polarized antenna pairs with ±45 degree
in most other places in the handbook, are drawn as classical polarizations. Two layers can then be efficiently multiplexed
beams, even though they often may be of generalized nature. even in the single propagation path case such as with free-
space channels, one on each polarization. Since signals with
orthogonal polarizations are well-isolated from each other in
a near free-space setting, there is little inter-layer interference
and SINR levels may be high. There is often significant isolation
between orthogonal polarizations also in non-line-of-sight
channels. Spatial multiplexing via exploitation of dual
polarizations is thus a key technique of great importance and
applicability. It also means that for each significant propagation
path in the simplified mental model, up to two layers can be
transmitted efficiently if dual-polarized antennas are employed
on both sides of the link.
MU-MIMO spatial multiplexing

N
How does it work?

• Several layers/beams to multiple UEs
• Different layers may go to different UEs
• UEs directionally well-separated: even classical beams
may have limited cross-talk
• UEs closely located: rely on base station side generalized
nullforming
• Several layers within same polarization possible even
with low angular spread
Capacity benefits in cells having high load with high SINR levels!
Spatial multiplexing for MU-MIMO is a way of increasing system This severely limits the nullforming capabilities on the UE side.
capacity by multiplexing several layers of data streams on the However, spatial multiplexing can nevertheless be supported
same time/frequency resources serving multiple users. also in this case if there is sufficient multi-path and the base
station uses generalized nullforming on its side to reduce the
MU-MIMO is easier to understand than SU-MIMO in the most inter-layer interference between the UEs.
common case when multiplexed UEs are sufficiently far apart
for them to be directionally well separated as seen from the base Similarly, as for SU-MIMO, MU-MIMO requires sufficiently high
station. The depicted obvious and intuitive mental model of how SINR. However, an additional requirement is that the traffic
the layers/beams towards different UEs have little cross-talk is load within the cell needs to be very high so that it is likely that
then applicable. The mental model holds if the channel angular at least two UEs in the same time slot wants to be served with
spread is sufficiently low. Even classical beamforming without data. Since high network load often means SINRs become lower
any base station side nullforming may under such favorable due to more inter-cell interference, the two requirements are
conditions be successfully used. somewhat in conflict with each other. This makes MU-MIMO a
capacity enhancing feature for a subset of the cells – those that
As multiplexed UEs come closer to each other it becomes more simultaneously have high load and many high SINR UEs.
challenging to support multiplexing of several layers. The
situation starts to resemble the SU-MIMO case with a UE with as
many antennas as the sum of antennas over all multiplexed UEs,
but with the crucial limitation that the multiplexed UEs cannot
process the received signals jointly across their combined set of
antennas.
No spatial multiplexing may lead

to inefficiency
Single layer Throughput

[bits/channel use]
One layer One data pipe High SNR Single layer

throughput
567
• All transmit power allocated to single layer • Throughput saturates for high SNR
• High SNR level for that single layer (flat throughput slope at high SNR)
An increase in already high SNR only gives a small increase in throughput!
To understand the basic principle behind the gain of spatial The upper right figure shows how the data rate throughput
multiplexing and the need for high SINR levels, it is instructive increases with an increasing SNR level ρ. It is observed how the
to consider a comparison between single and dual layer slope of the throughput curve changes from being steep at low
transmission. First, consider as depicted at upper right a link SNR levels while flattening out at high SNR levels. In particular,
where a single layer is transmitted creating a single data pipe. the throughput saturates for high SNR meaning that even if the
The modulation symbol s is transmitted over an additive white SNR would be increased further from an already high SNR level,
Gaussian noise (AWGN) channel to produce a noise impaired there is very little additional throughput gain. Beamforming
received symbol ŝ that experiences an SNR level of ρ. The total increases the SNR and hence as the beamforming gain improves
transmitted power is P and will subsequently be kept constant of a Massive MIMO system with larger and larger antenna arrays,
regardless of the number of layers. This is a highly simplified the pay-off in terms of increased throughput becomes smaller
model of a communication link but will nevertheless serve the and smaller.
purpose of illustrating the principal differences between single-
layer and multi-layer spatial multiplexing. It would be unfortunate to not be able to benefit from high SNR
levels to substantially improve user experience via better data
rate throughput. As will be demonstrated on the next page,
spatial multiplexing is a way of exploiting high SNR levels for
obtaining data rate improvements.
Why spatial multiplexing provides

gain at high SNR N
Example: single versus dual layers

1 =
Transmit power shared by two layers
Dual layers Two data pipes Half SNR 2 =
Throughput
[bits/channel use] dual layer
throughput
layer 1 throughput layer 2 throughput spatial

multiplexing
gain
single layer
throughput
SNR
Net gain: More data pipes compensate for lower SNR per data pipe!
Consider as depicted a link with two layers, where each layer As previously explained, the data-rate for a single-layer (black
k transmits a modulation symbol skover an additive white curve) is seen to saturate as its SNR=ρ increases—trying to
Gaussian noise (AWGN) channel to produce a noise impaired increase an already high SNR even further does not pay off much
received symbol ŝk that experiences an SNR of ρk. For simplicity, in term of data-rate increase. For the two-layer case it is seen
cross-talk between the two parallel AWGN channels is neglected how the red and green layers each get half the SNR=ρk=ρ/2
and thus the two layers are transmitted on two independent via power sharing and as expected the throughput Ck for an
“data pipes”. The total transmitted power P is the same as in the individual layer is less than in the single layer case. However,
single-layer case on the preceding page. Hence, the transmit the sum data rate Ctot=C1+C2 is larger! This since the individual
power for each layer is P/r, where r is the number of transmitted layers operate on a steeper lower SNR part of the throughput
layers. In other words, the transmit power per layer decreases as curve. In essence, a spatial multiplexing gain is achieved by
the number of layers increases. In the above example it is seen sharing a “too high SNR” among more layers.
how the single layer SNR of ρ is halved to ρk=ρ/2 for each layer
in the two-layer case. Thus, the SNR, and hence the data-rate, for This simple special case illustrates the basic principle explaining
each layer is lower than in the single layer case. There is however how spatial multiplexing achieves its benefits and why it thrives
two layers instead of only one, so it is conceivable that the sum on high SNR. The AWGN model of the link is obviously over-
data rate over the two layers could be larger than the data-rate simplified and idealistic, but the main qualitative conclusion still
in the single-layer case. applies to more realistic models.
To see that indeed the reduction in per layer data-rate is more In a more realistic setting inter-layer interference would act to
than compensated for by the increase in number of parallel decrease the per layer S(I)NR compared with the analysis above.
layers/data-pipes, consider the bottom graphs. The two left This means that using too many layers in relation to the SNR
graphs show how data-rate depends on the SNR for each of the may incur a loss in sum data-rate, instead of as above always
two layers separately. The throughput for layer 1 and layer 2 are a gain. Carefully adapting the number of layers to the channel
identical and are summed together to produce the dual layer conditions at hand thus becomes important.
throughput as illustrated in the right-most graph.
Number of useful layers decrease

as cell edge is approached No chang
4 layers 3 layers 2 layers 1 layer
The previous two pages explained why spatial multiplexing can Efficient wireless systems perform dynamic adaption of the
give a throughput gain. For specific channel conditions, there is number of layers depending on many factors, including the
an optimal number of layers to use, and that number tends to channel conditions and number of UEs that have data in the
increase with the SINR level. buffer at a particular time instant. Inter-cell interference acts as
adding noise and hence the SINR levels in a cell tends to vary
In a cell, SINR tends to decrease with distance to the base with the load of the neighboring cells. As the load in neighboring
station. Consequently, the optimal number of layers is decreasing cells increases, the optimal number of layers to use in the cell at a
as the cell edge is approached. Using many layers is beneficial particular location is reduced.
close to cell center while fewer, or only one layer, is used at
cell-edge.
Determining transmit beamformer/precoder

from channel state information (CSI) No
Codebook based feedback Reciprocity

• UE measures the channel based on a base • Base station measures uplink channel based on a UE
station transmission transmission
• Measurements è UE selects beamformer from a codebook • DL channel properties inferred from uplink channel
• Indication of recommended beamformer sent to • TDD è UL & DL channels are the same (reciprocity)
base station
• Base station uses DL channel for computing a beamformer
Beamformer indication Reciprocity signal
Channel knowledge, or so-called channel state information Reciprocity

(CSI), is needed to know where to beamform. There are two The reciprocity approach exploits that the channels in both
different approaches to CSI acquisition for transmission the forward and reverse directions are more or less similar. The
purposes. transmitting base station can therefore do channel estimation
from reverse link measurements and use that to determine a
Codebook based feedback precoder in the forward link towards the UE:
In codebook-based feedback, the codebook corresponds
to a finite and countable set of predetermined precoders/ 1. Base station estimates the uplink channel based on
beamformers. Precoding is a popular and more specific term for a UE transmission of known reference signals
beamforming in the transmit direction. It is the UE that, based on
channel estimates, determines a precoder from a codebook and 2. Base station uses uplink channel estimate to infer
thereafter feeds that precoder back to the base station so that properties of the downlink channel
the base station can use that precoder recommendation when
3. Base station uses downlink channel to determine
determining how to precode its transmitted signals towards the
a precoder to use in its transmission towards the UE
UE:
Reciprocity based techniques tends to lead to beam patterns
1. UE estimates the channel based on receiving known
similar to generalized beamforming.
reference signals from the base station.
2. UE selects a precoder matching the estimated channel For TDD, uplink and downlink are on the same frequency and
properties from a codebook. hence the corresponding channels are identical. The inferred
and actual downlink channels thus match very well, including
3. UE feeds back the selected precoder(s) as a recommendation their fast fading properties. Reciprocity also works for FDD but
to the base station. in a more limited way. Uplink and downlink are then on different
4. The base station uses the recommended precoder in its frequencies so the inferred downlink channel may capture
transmission towards the UE, or uses it as one source of input reciprocal large-scale (/long-term) channel properties such as
in proprietarily determining a precoder for the transmission. propagation directions and path loss, but not the fast fading.
In either case, reciprocity requires base station tx/rx antenna
Codebooks are often based on grid of beams (GoB) precoders calibration.
and as such codebook-based feedback tends to be related to
classical beamforming.
Where to use different CSI

acquisition methods
Codebook b
Fast fading CSI out of coverage
Fast fading CSI coverage

1. TDD with short-term reciprocity Long-term CSI coverage
2. Advanced codebook-based feedback 1. Long-term reciprocity (TDD/FDD)
2. Conventional codebook-based
feedback
CSI can come in different detail levels. In particular, there can be Reciprocity for TDD is one way of obtaining short-term CSI.
CSI that includes short-term (/small-scale) channel properties Advanced codebook-based feedback, as discussed in Chapter
or CSI that only contains long-term (/large-scale) channel 4 [TP4, p. 52], is another. More conventional codebook-based
properties. Reciprocity for TDD capturing fast-fading properties feedback primarily focuses on providing long-term CSI. Similarly,
is an example of the former, while reciprocity for FDD is an reciprocity for FDD is also limited to providing long-term CSI.
example of the latter. However, for TDD, using long-term CSI is more of a voluntary
option that may be beneficial where coverage is bad.
Including short-term properties provides more details
(constituting more “information”) and hence puts stronger These observations lead to the conclusion that it can be
demands on a good signal level to be able to reliably extract a beneficial for the base station to switch CSI acquisition scheme
larger amount of CSI. Thus, schemes relying on short-term CSI depending on where in the cell the UE of interest is. The farther
usually have smaller coverage than schemes that only rely on out in the cell a UE is, the more likely it becomes that a simpler
long-term CSI. On the other hand, performance is usually better but more robust, long-term CSI scheme is preferable.
for short-term-based schemes when coverage is sufficient to
support them.
Partial channel reciprocity limits

performance Re
Full channel reciprocity Partial channel reciprocity

(half channel reciprocity)
Full channel
Full channel reciprocity
reciprocity Partial
Partial channel
channel reciprocity
reciprocity
(half channel
(half channel reciprocity)
reciprocity)
TX TX TX
TX
− − 𝒉𝒉'( −− RX − − 𝒉𝒉'( −−
UE UE
RX RX
UE UE
RX
𝑯𝑯"# = 𝑯𝑯"# =
− − 𝒉𝒉() −− TX TX
− − 𝟎𝟎( −− RX RX
RX RX
Ideal for reciprocity-based beamforming! Harmful for reciprocity-based beamforming!
In NR and in LTE, UEs may have more receive antennas than For LTE, the UE norm is to have two receive antennas and one
transmit antennas. This presents a problem for reciprocity non-switchable transmit antenna. Suffering from half channel
because uplink channel measurements will then only provide reciprocity is thus common. In NR, the situation is less clear
part of the whole downlink channel, so-called partial channel and it will take time to be able to judge what the majority of
reciprocity. The figure above illustrates the difference between UEs support. In the present still early phase of NR, it seems UEs
full and partial (half) channel reciprocity. Having only parts of the commonly have four receive antennas and possibility to switch
channel for determining downlink precoders in the base station transmit antenna(s) for sounding reference signal purposes. This
significantly impairs the performance since the unknown part of avoids the partial channel reciprocity problem for the present UE
the channel prevents an effective precoder to be formed for all of population, but this may change as the UE population inevitably
the UE receive antennas. It may even result in reciprocity ceasing shifts from high end to lower end UEs.
to be a competitive technology.
While it may be prohibitive for many UEs to have the same

number of radio chains for transmit as for receive, some UEs
support transmit antenna switching so that which of its antennas
are used for transmission can be dynamically switched from one
time instance to another. This allows the sounding reference
signals commonly used for reciprocity measurements to
alternate between transmitting on the various UE antennas. Full
channel reciprocity can in this way be achieved despite having
fewer radio chains (even only one) in the transmit direction.
Potential drawbacks include increased sounding overhead
or time to acquire all channel measurements as well as some
insertion loss due to the switch.
MU-MIMO
—realistic expectations? No chang
Performance:
Easy to get too high expectations…
Industry demos
• Single cell
• Full buffer traffic
• All connected UEs
assumed dynamically active
• Real-life practical issues often
neglected!
MU-MIMO is clearly an important technique for boosting the In industry demos, it is very common to use only a single cell
capacity in highly loaded cells. The technique has been seen as and therein manually place, often in MU-MIMO favorable high
the essence of Massive MIMO ever since its introduction in the SNR locations, a bunch of UEs. This is illustrated above and
literature. Extraordinary performance gains over SU-MIMO have neglects that in real networks there is often substantial inter-cell
been demonstrated both in the academic literature as well as in interference that will lower the SINR levels and thus the gains of
industry demos. Expectations have therefore often been set very MU-MIMO with many layers over fewer layers SU-MIMO. Also,
high and it is easy to get the impression that adding more and the data traffic is assumed to be full-buffer, meaning that all UEs
more layers is always key to good performance and would offer want infinite amount of data in every time slot. In reality, the
a factor of 10 or more in performance boost. In contrast, realistic data is packet based and often very bursty with significant silent
simulations and trial results show good gains of MU-MIMO periods between the packets. This reduces the likelihood that
in certain scenarios, but the gains are more on the order of a many UEs want data in the same time instance and thus limiting
factor of 1.5 or somewhat more (c.f. generalized vs. classic the amount of MU-MIMO layers needed compared with the
beamforming for 64 T in [TP9, p. 161]). unrealistic full buffer assumption. Also, in real networks there is
often a strong tendency that the set of UEs that want data in the
There is thus a risk that the expectations of MU-MIMO are same time slot is dynamically changing over time, making it hard
unrealistically high. There is in fact an abundance of differences to provide the necessary accurate channel knowledge in advance
between real-life commercial wireless networks and what of the scheduling of the UEs.
is typically being assumed in industry demos and academic
literature. These differences unfortunately tend to all artificially
boost the relative gain of MU-MIMO over SU-MIMO compared
with what will be the case in realistic settings. This and the next
page will provide more details on why there is a tendency to get
unrealistic MU-MIMO gains in industry demos as well as in the
academic literature.
MU-MIMO
—realistic expectations? N
Performance:
Easy to get too high expectations…
Academia
• An extreme channel model: IID Rayleigh fading
• SU-MIMO crippled UE: single- antenna or layer UE
• Full channel reciprocity: #TX = #RX antennas for UE
• Optimistic quality of reciprocity-based CSI
• Narrowband/SINR/frequency
• Relying on UE Tx inter-slot phase coherency
• Real-life practical issues often neglected!
The academic literature in general strives for closed-form The assumption of a single receive/transmit antenna at the
analysis. As such, several simplifying assumptions are often UE also means that the so-called partial channel reciprocity
made as discussed below. problem does not exist. This greatly benefits reciprocity based
schemes since no channel information is missing and MU-MIMO
It is very common to use a Rayleigh fading channel model in particular relies on accurate channel knowledge, while
with spatially uncorrelated fading. This model has no notion of SU-MIMO is less demanding when it comes to quality of channel
“direction”, all channel directions are equally likely. In contrast, knowledge.
real channels have a significant spatial correlation and have
widely different characteristics vertically and horizontally. Many The quality of reciprocity-based channel estimates may also be
UEs tend to be confined in a small angular range in the vertical too optimistic as often the assumed bandwidth is small so uplink
domain with high spatial correlation and this seriously limits the SINR per frequency becomes high. There may also be filtering
multiplexing potential for MU-MIMO. of channel estimates across time-slots, something that requires
that UEs support phase coherency over time, which current
Academic studies commonly assume all UEs have only a single commercial UEs do not.
receive/transmit antenna, which restricts SU-MIMO to only a
single layer and removes possibility to do inter-cell interference
rejection in the UE. This cripples in particular the SU-MIMO
performance, thereby artificially increasing MU-MIMO over
SU-MIMO gains.
Summary
Summary Second
note pa
• A beam may not look like a beam
• Beamforming for improving signal levels

is the main component in Massive MIMO
performance
• MU-MIMO important capacity enhancer

but its gains are often exaggerated Performance:
This chapter addressed miscellaneous technology components increase in SNR levels is also a prerequisite for techniques such
essential to massive MIMO in greater detail than in the main as MU-MIMO to thrive.
story. There are many conclusions and findings in the chapter.
The three most important take-aways can be summarized as: •
• MU-MIMO is surely an important capacity enhancement

• A beam may not look like a beam, i.e., like a single directed
technique for those cells having very high load. It is possible
entity as implied by intuition. It may in fact look like an arbitrary
to demonstrate extraordinary performance gains on the order
blob when there is significant angular spread in the channel
of a magnitude or so under very specific artificial conditions.
and/or MU-MIMO is being used. Yet for simplicity, and to focus
Academic literature and artificial industry demos have
on the concept of beamforming, illustrations typically show
consequently set very high expectations and massive MIMO
very well-behaved beams with a strong dominating main lobe
has for this reason been thought to be synonymous with
(also this handbook does that). Confusion arises when this
MU-MIMO. The specific conditions needed for achieving these
leads to a belief that this is how a beam is supposed to look like
kind of gains are however rarely found in realistic network
while in reality the exact pattern of the beam is often much less
settings and thus MU-MIMO capacity gains in practice are
relevant.
considerably less, although still good and often around a
• respectable 30-50%.
• Beamforming for improving signal levels is the main

component in massive MIMO performance. The large
beamforming gain from large antenna arrays and use of UE
specific beamforming ensures great coverage and facilitating
reuse of a site grid for higher frequency bands. The resulting
4. 3GPP –physical layer

support for Massive MIMO
4. 3GPP — physical layer support for Massive MIMO

Ericsson | The Massive MIMO Handbook — Technology Primer 4. 3GPP — physical layer support for Massive MIMO 45
3GPP Physical layer components

—introduction N
The purpose of the following

pages is to give some background
of relevant parts of the 3GPP TM
physical layer specifications
The purpose of this chapter is to and types including MIMO codebooks
• Give some background of relevant parts of the 3GPP physical • This is a key area for Massive MIMO performance and
layer specifications. understanding the different features
• Assist the reader to understand terms and concepts for the • Sounding Reference Signal (SRS) and antenna switching
sections feature solutions and network performance.
• For reciprocity based Massive MIMO, the uplink RS is
The content and motivation/relevance for Massive MIMO: important, and the basics is covered here
• 3GPP physical channels • Periodic, semi-persistent and aperiodic behavior of CSI

reporting, SRS, CSI-RS
• Key to understand the physical layer of NR
• The 3GPP specifications allow a large flexibility to configure
• Carrier aggregation (CA) and dual connectivity (DC) time behavior of measurements and reporting, and this
provides a brief overview
• The combination of Massive MIMO and carrier aggregation
is common, and it is important to at least grasp the For a more extensive description of the 3GPP physical layer
fundamentals of carrier aggregation. standard related to AAS (Massive MIMO), refer to [1].
• Cell defining signals and control channels
• This part is done differently in NR compared to LTE due to

introduction of beam based initial access by Massive MIMO
• Channel State Information (CSI) feedback principles

Summary of the physical channels in NR
gNB
PD C
CH,
PD S
CH
PUC
CH,
PUS
CH,
PRA UE
CH
SS/PBCH
In order to understand the different physical channels in NR, The three uplink channels in NR are:
which are combined with the Massive MIMO technology
components in [TP3], an overview is given here. See also the • Physical uplink shared channel (PUSCH) carries the UE uplink
3GPP specifications TS 38.211, TS 38.213 and TS 38.214 (3gpp. data and/or aperiodic CSI feedback
org).
• Physical uplink control channel (PUCCH) carries the uplink
The three downlink channels in NR are: control Information (UCI)
• Physical random access channel (PRACH) which is used for

• Physical broadcast channel (PBCH) is a broadcasted channel
random access and to adjust uplink timing
(potentially received by all UEs) which carry small but essential
part of system information needed to access the cell. The PBCH is transmitted adjacent to the synchronization signals
(SS) and carries a few bits with info on how to access the cell at
• Physical downlink control channel (PDCCH) carries the
initial access. The PBCH+SS (known as an SSB) covers the whole
downlink control information (DCI), used for scheduling PDSCH
cell, or a portion of the cell, in case beamformed SS/PBCH is
and PUSCH etc.
used.
• Physical downlink shared channel (PDSCH) carries the
downlink data The channels have different coverage (in some cases depending
on the information payload and used code rate) and these
differences needs to be taken into account when planning the
deployment of the network.
Note that NR also support side-link and other recent features,

for which additional physical channel has been added in later NR
specification releases.

Used terminology in carrier aggregation
Master Cell Group (MCG) Secondary Cell Group (SCG)

DL PCell SCell SCell DL PSCell SCell
SCell
(SpCell) (SpCell)
UL UL
Carrier aggregation of 4 cells Carrier aggregation of 2 cells
Dual connectivity using two cell groups
Band A (e.g. 3.5 GHz) Band B (e.g. mmWave)
!"#$$
PCell SCell Contiguous Intra-band CA
PCell SCell Non-contiguous Intra-band CA
PCell SCell Inter-band CA
Carrier aggregation (CA) is an important feature for Massive A Secondary Cell (SCell) may not have an uplink (for a given
MIMO as CSI reporting can be configured to carrier with better UE) and may or may not be configured with SS/PBCH. If not
uplink coverage. In Chapter 3, Solutions, the Beamforming/ configured, the UE use SSB of another cell.
carrier aggregation mode switching is one example where carrier
aggregation is combined with Massive MIMO. Here, the basics of • May have a different SCS and duplex mode compared to the
carrier aggregation is introduced, see TS 36.300 and TS 36.331 Pcell
(3gpp.org) for the relevant specifications.
• For example, in TDD, an SCell can be downlink only and the
In NR, “cell” is used to denote a carrier in carrier aggregation and UE cannot transmit SRS for reciprocity-based operation. If
up to 16 such serving cells can be configured to the UE using supported by the UE, SRS Carrier Switching can be used in this
non-overlapping frequency bandwidths. A serving cell can either case.
be a special cell (SpCell) or a secondary cell (SCell). Furthermore, • An SCell can rapidly be activated or deactivated depending on
the SpCell is either a Primary Cell (PCell) or a Primary SCG Cell the need (latency is a few ms if MAC CS is used (Rel.15) or less if
(PSCell). Frequency bands are defined and there is a distinction DCI is used (Rel.16))
of intra-band CA (e.g. among carriers at 3.5 GHz carrier
frequency) and inter-band CA (e.g. between carriers at 3.5 GHz There is also a possibility to configure a supplementary uplink
and 28 GHz carrier frequencies). (SUL), where a downlink carrier have two uplink carriers. If
dual connectivity is configured, a master cell group (MCG) and
In addition, in intra-band CA, there is a distinction between a secondary cell group (SCG) is defined. Both the MCG and SCG
contiguous and non-contiguous CA, where in the latter, the has a “PCell”, and it is defined as PSCell for the SCG. The PSCell
carriers are not adjacent in frequency. handles random access and handle RLM just as the PCell in the
MCG.
A Primary Cell (PCell), is a SpCell/PSCell and it always has
an uplink which also carries uplink control information (UCI)
feedback (e.g. HARQ-ACK,CSI) for the Pcell and by default, also
UCI for all SCells unless configured differently
• PCell handles initial access/receives the Random Access

Channel, handle Radio Link Monitoring (RLM)
• The cell that is the PCell can be reconfigured for the UE, using
the handover procedure which takes about 40 ms

Overview of NR synchronization signals

No chan
A burst of L=4 sync signals transmitted in L=4 different directions
SCS (kHz) Frequency range Lmax

15 ≤3 GHz 4
15 3 GHz to 7.125 GHz 8
30 ≤3 GHz (paired 8
SSB Time spectrum)
Slot 30 ≤2.4 GHz (unpaired 4
spectrum)
SSB burst set (5 ms)
SSB Periodicity (20 ms) for initial cell search 30 2.4 GHz to 7.125 GHz 8
(unpaired spectrum)
120 or 240 24.25 GHz to 52.6 64
Lmax is the maximum number of GHz
supported SSB beams per frequency 120, 480 52.6 GHz to 71 GHz 64
range and subcarrier spacing or 960
The cell-defining signal is the signal or set of signals the UE A SIB1 is transmitted in each beam as the associated SSB, but
detects from an SpCell and uses as the basis to begin access not necessarily adjacent in time to the SSB but at specified point
procedures to establish a connection to a cell. If these cannot be in time relative to an SSB.
reliably received, then the UE is outside that cell area.
To conclude, initial access signals can flexible be configured in
In NR, the cell defining signals are the synchronization signals NR, as opposed to LTE (which would correspond to Lmax =1, i.e.
(SS) together with the broadcast channel (PBCH) and the system a non-beam based initial access system). Different operators or
information block 1 (SIB1) which is carried on PDSCH. network vendor may therefore approach this differently, e.g. the
selection of number of cell defining signals L. There are pros and
The cell defining signals are specified differently in NR compared cons to select a large or a small (e.g. L=1) value such as larger
to LTE since a cell can be configured with multiple cell defining L implies more SSB overhead but better match between data
signals, which are time multiplexed and transmitted in different beam and cell defining beams.
beam directions. The maximum number of such cell defining
signals Lmax depends on the subcarrier spacing (SCS) and
frequency range, see the table.
The SS/PBCH (SSB) duration is 4 OFDM symbols and 20 RB

wide and it contains the primary and secondary synchronization
sequence (PSS/SSS) and the physical broadcast channel
(PBCH). The SSB period is 20 ms and are localized in a burst
set of duration 5 ms in the beginning of the 20 ms period, see
the example figure of L=4 SSB and thus four different beam
directions. Note that the burst set length and periodicity (5 ms
and 20 ms) are independent of the subcarrier spacing. With
increased subcarrier spacing, there are more slots within the
burst set duration and thus room for a larger L.

What is Channel State Information (CSI)?

C
Massive MIMO antenna array

with 64 steerable antenna
subarrays
Multipath channel
Example of the set of beams in
the Codebook. Each beam is formed
by coherently combined transmission
using all 64 subarrays
The information of the multipath propagation channel between To reduce CSI reporting overhead and complexity, a codebook
gNB transmitter and UE receiver is in 3GPP called channel state has been defined in 3GPP which specifies a small but highly
information (CSI). It is conveyed to the gNB in a CSI report and efficient subset of possible phase settings per antenna subarray.
used by gNB to determine how the Massive MIMO antenna array 3GPP has defined such codebooks under some assumption of
should transmit to maximize e.g. the received signal power at UE. typical Massive MIMO antenna array structures (such as a planar
For Massive MIMO solutions discussed in Chapter 3, the concept array with element spacing in the order of a wavelength), and
of CSI is central, as it is crucial for good MU-MIMO performance typical mobile communication radio channels,
but also for SU-MIMO. See TS 38.214 (3gpp.org) for the related
specification. Note that a beam is the results of transmission from all subarrays,
with carefully adjusted phase shifters to coherently combine the
To acquire CSI at the gNB for a given UE, the gNB can signals from these subarrays in a certain desired direction.
alternatively use reciprocity, i.e. measure on an uplink
transmission from the UE such as SRS or DMRS transmission. The takeaway is that the channel is multi-dimensional and to
quantize it for reporting would cost too much in CSI reporting
The Massive MIMO antenna array use a large number of overhead. Instead, a carefully designed codebook of a set of likely
controllable antenna elements, using phase shifters (and transmission hypotheses (e.g. beams) is used which reduces the
sometimes also amplitude shifters). The objective of acquiring overhead. Alternatively, the gNB can use uplink measurements
CSI at the transmitter, is to improve the performance of the and rely on reciprocity to acquire CSI.
downlink transmission by utilizing information about the channel
between each antenna element (e.g. each subarray or antenna
port) of the gNB and each receiver antenna element of the UE.

The UE measures on an equivalent,

“logical” antenna array No chan
Mapping each CSI-

RS port to an
antenna subarray
Mapping each CSI-RS
port toeach
Mapping an antenna
CSI- subarray
RS port to an
antenna subarray
A ”logical” antenna array defined by Physical antenna Set of possible beams as

A “Logical”
the CSI-RS antennato array
ports and configured UE (not observable by UE)
Set of possible beams
defined by the codebook
defined by the 16 CSI-RS Physical antenna
as defined by the
ports and configured to (not observable by
codebook
A “Logical”
UE antenna array UE)
Set of possible beams
defined by the 16 CSI-RS Physical antenna
as defined by the
ports and configured to (not observable by
codebook
UE
In codebook based operation, See Chapter 3 and TS 38.214
UE) From the N CSI-RS ports, the codebook defines a 2D grid of
(3gpp.org), the UE measures the downlink channel using CSI-RS beams (GoB), where 4*N1 beams are in the first dimension
and finds the best possible precoder W and transmission rank and 4*N2 beams are defined in the second dimension, in total
for a hypothetical PDSCH transmission. It should also report 16*N1*N2 beams. See the figure to the right. The use of 4 times
a channel quality index (CQI), which is the modulation and the number of beams compared to the number of ports of that
coding scheme that gives the highest throughput given a target dimension, is denoted spatial oversampling, as more beams
transmission error rate (e.g. 10%). Note that in codebook based than just the set of orthogonal beams can be selected. This
operation, the beams from which a selection is done are virtual reduces the losses as beams in between orthogonal beams can
and exists only in the UE based band processing, the actual be selected.
CSI-RS transmission does not represent multiple beams. This has
the advantage that a large number of beam hypotheses can be A beam is created using a linear phase progression across the
evaluated from a single CSI-RS measurement. spatially separated antenna ports, per polarization, and is
in 3GPP defined using the discrete Fourier transform (DFT),
Assume the AAS consists of 8 subarrays which are arranged which has exactly this property. The CSI report also contains
as 2x4 (see the middle figure). Each subarray has 4x1 column co-phasing factors, for example if rank 1 is selected, the UE
of dual polarized antenna elements, hence in total 2x4x4x2 = reports how to co-phase the two polarizations within a selected
64 “antennas”. In the implementation it is chosen to use two beam. The co-phasing can change rapidly across the bandwidth
CSI-RS ports per subarray, one per polarization. Hence, the and can thus be configured to be reported per subband, while the
UE is a configured MIMO codebook with N1=2 rows and N2=4 beam selection is the same across the whole bandwidth.
columns in the antenna port layout (left figure). The UE is thus
configured a CSI-RS resource with 2*N1*N2=16 CSI-RS ports.
Therefore, from the UE perspective, the AAS array appears as a
2x4 antenna array of dual polarized antenna elements and the
implementation using subarrays is not “visible” to the UE.

Type I CSI (coarse channel report)

N
Ignored (weaker) beam
Ignored (weaker) beam
Sele
cted
(stro
nges
t) be
a m
Ignored (weaker) beam UE selects strongest beam in the

codebook for a Type I CSI report
For CSI reporting, either Type I or Type II codebook based Alternatively, the gNB is transmitting a single CSI-RS resource
reporting can be configured to the UE. See Chapter 3 and TS of N ports, where each such port covers the whole sector, and
38.214 (3gpp.org). The figure shows an illustration for the the multiple beams are the created virtually in the UE by using a
example of rank 1 CSI reporting for Type I (and only one of codebook of pre-defined beams as described in previous pages.
the polarizations is shown for simplicity). In Type I, the main This is denoted antenna-based CSI reporting.
principle is that only the strongest reflection in the channel is
selected and reported as CSI to the gNB, the weaker ones are In Type I reporting, the antenna-based reporting is used, and the
ignored. This type of CSI is often sufficient for low rank SU-MIMO UE evaluate the whole set of beams in the codebook and selects
transmission, since the gNB only need to know where to focus all one preferred beam. For higher rank, multiple beams are selected.
it’s transmission power to reach the UE. The Type I CSI report has The selection thus corresponds to a preferred precoding matrix W
low payload, compared to the richer Type II report. that is fed back to the network using the PMI field in the report.
The PMI field also contains co-phasing of the polarizations
In both Type I and II cases, a basis set of dual polarized beams (note that the figure assumes a single polarization for simplicity
mapped to vertical and horizontal, are used for feedback of illustration) and this can be configured to be reported per
reporting. Either the gNB is creating the set of beams by actually subband. As this report only report the most preferred direction,
transmitting multiple CSI-RS by using antenna virtualization V. it works well for SU-MIMO scheduling and transmissions. The CSI
One CSI-RS of 2 ports (one port per polarization) is in this case reporting overhead is low.
transmitted in each beam and the UE is selecting the preferred
beam for reporting. This is denoted beam-based reporting.

Type II CSI (rich channel report)

No chan
ker) beam
Selected (wea
ker) beam
Selected (wea
Sele
cted
(stro
nges
t) be
a m
UE selects four strongest beams

Selected (weaker) beam out of N, and report together with
amplitude and phase
The figure shows an illustration for the example of rank 1 CSI In Rel.16, an enhancement was introduced to the Type II
reporting for Type II CSI (and only one of the polarizations is codebooks. The fact that the co-phasing per subband is
shown for simplicity). The principle is to feed back a rich channel correlated across adjacent subbands was utilized in order to
report, aimed for MU-MIMO scheduling. It is important to reduce the CSI reporting overhead. Thereby, the performance
know how a transmission towards one UE interferes another, improves, or the overhead reduce, depending on the chosen
co-scheduled, UE. Therefore, not only the strongest beam is configuration.
reported but also weaker beams, as they are utilized to compute
this interference. A drawback of Type II is the larger reporting In Rel.17, a further enhancement was introduced for the Type II
overhead. codebook, targeting mainly FDD deployment. It is observed that
due to the FDD duplex distance, the phase information is not
In Type II reporting, the UE evaluates the whole set of beams in reciprocal between uplink and carrier frequencies, but the angles
the codebook and selects multiple preferred beams. It scales the and delays maintain reciprocity despite the duplex distance.
power of each beam according to 4 quantization levels, based Hence, the gNB can measure angles of arrivals and delays of
on the CSI-RS measurement and computes how these beams the channel components and transmit beams in the downlink
should be co-phased using a QPSK or an 8-PSK alphabet. The which are pre-compensated with respect to delays (and beam
purpose is to report multiple beams and for each, both amplitude directions). This will reduce the UE complexity significantly, while
values and co-phasing. This gives a good representation of the more computations are needed on the gNB side.
whole channel, not only for the main direction as in the Type
I reporting. The overhead depends on whether the amplitude
report and co-phasing is subband or wideband. Also, the number
of beams the UE shall select can be configured to the UE. When
performing MU-MIMO scheduling, knowledge of the whole
channel matters, not only the main direction, since this gives the
network a possibility to control the interference that is created
towards co-scheduled UEs.

Selection of precoder W utilizes the CSI

M
a
Effective channel for PDSCH demodulation
Effective channel for CSI measurements
CSI-RS port 3000

DMRS port 1000
CSI-RS port 3001
Code word
Antenna Channel H Receiver
to layer Precoder W
virtualization V
mapping
DMRS port CSI-RS

1001 W port 3002
CSI-RS port 3003 Codebook based: CSI feedback including W
Scheduler
Reciprocity based: CSI acquisition based on SRS
The figure illustrates a downlink transmission chain for a MIMO In case of reciprocity based operation, the UE transmits sounding
system that is the foundation of the solutions described in reference signals (SRS) in the uplink, so that the gNB can
Chapter 3 and is the assumed model used in 3GPP specifications estimate the uplink channel. In TDD, the uplink and downlink
of single TRP transmission of data to a UE. The encoded set channel H are reciprocal in case of full reciprocity. See the details
of bits (code word) is mapped to layers before precoding and on SRS antenna switching on a following page on how 3GPP
transmission. The OFDM block that converts from frequency to have tried to address the partial channel reciprocity issue.
time domain is placed before or after the antenna virtualization
(not shown). When transmitting PDSCH, the stream of modulated symbols
(known as a code word), is mapped to layers (i.e. the rank of the
3GPP defines “antenna ports” by inserting reference signals transmission to the UE). One demodulation reference signal
at a point in the transmission chain. In this figure, four CSI-RS (DMRS port) is transmitted, per layer. The DMRS is added before
are inserted, but these pass through an antenna virtualization the precoder, so that the transmission will pass through the
matrix V, hence, the UE measures the effective channel for combined channel HVW before being received in the UE receiver
CSI measurement HV using CSI-RS. It is therefore up to gNB antennas. Hence, when UE measures the channel using DMRS, it
implementation how to choose V, it can be used to beamform measures the effect of the combined, effective channel for PDSCH
in some direction if desired. The UE is not aware of how V demodulation, on the transmitted modulated symbols. The
is selected. In mmwave operation, it is common that V is precoder W is thus also part of the effective channel for PDSCH
implemented using analog phase shifters, i.e. a time domain demodulation. Therefore, for data transmission, gNB can choose
analog signal processing operation. In addition, each physical arbitrary W, and what the UE feeds back in a CSI report is just a
antenna shown here may actually represent the transmission recommendation.
from a subarray of physical antennas.
In case of codebook based operation, the UE estimates a rank

and a preferred precoder W (PMI + rank feedback) in the CSI
feedback report to the gNB. The network can configure the W to
be a wideband report (one precoder for the whole bandwidth), or
subband W reporting, which consumes more reporting overhead
but gives increased channel information as the channel is
frequency dependent, in general.

SRS antenna switching options

for 2 RX UE No chan
1T2R
UE implementation that
requires antenna
switching. Guard symbol
needed for antenna
switching.
2T2R
UE implementation
does not require
antenna switching
Sounding reference signals (SRS) can be configured to be The UE can be configured to intermittently transmit SRS in such
transmitted from the UE and is important in reciprocity-based uplink carrier (which is an uplink carrier that cannot be used for
operation, particularly for MU-MIMO scheduling. A UE may transmitting data or control) so that reciprocity transmission
have fewer TX antennas than RX antennas and in this case SRS can be used for the corresponding downlink. However, when
antenna switching can be utilized so that the gNB acquires CSI performing such temporary switch to another carrier, the
for the full MIMO channel (full channel reciprocity) and not only transmission on the default carrier is paused.
partial channel.
In Rel.15, the SRS (including guard symbol) can be transmitted
A terminology for antenna switching is introduced where 1T2R in one or more symbols in any of the last 6 symbols in the slot
means that the UE have 1 TX and 2 RX antennas and support and the SRS bandwidth is configurable, while in Rel.16, the
antenna switching. The UE will report 1T2R as a capability to up to 6 SRS symbols can be transmitted in any set of adjacent
the network in this case. If the UE has full reciprocity capability symbols of the slot. Rel.17 further enhanced this by removing the
the UE can in this case report 2T2R and there is no need to use restriction of 6 symbols.
antenna switching in this case. See the figures for examples
where also the TDD switch is illustrated. It should also be noted As the total output power is limited, the wider the SRS bandwidth
that antenna switching is costly in overhead in that it (currently) the worse is the coverage. Interference mitigating and coverage
requires an empty Guard Symbol to allow for the UE to perform enhancing techniques such as frequency hopping (FH) and
the antenna switch. For UEs with 4, 6 or 8 RX antennas, there are repetition with a factor of 2 or 4 can be also configured. In FH, the
similar switching schemes, e.g. 2T4R, 2T6R, 2T8R, 4T8R, … and SRS is transmitted in a different part of the bandwidth in each
it could in those cases be even more guard symbols as there are symbol, while in repetition, the same part is used in each symbol.
more antenna switches needed.
The SRS can be configured to be transmitted on every 2nd or
In carrier aggregation, not all downlink carriers have an every 4th subcarrier (known as “combs”) which defines up to four
associated uplink carrier, and to be able to transmit SRS also in orthogonal resources for SRS. In Rel.17, every 8th subcarrier is
such carrier without an uplink, SRS carrier switching has been additionally supported.
introduced.

CSI measurements and reporting can be

configured using three approaches
RRC config (PDSCH)

Three possible mechanisms available to the network operator, Example of SRS,
MAC CE (PDSCH) CSI-RS transmission
to trade-off overhead, flexibility, energy consumption, etc.
or a CSI report
DCI (PDCCH)
Configure (RRC)
Periodic
Configure (RRC) Activate (MAC CE) De-activate (MAC CE) Activate (MAC CE)
Semi persistent
Configure (RRC) Trigger (DCI) Trigger Trigger Trigger

Aperiodic
NR support three mechanisms to enable or disable measurement Semi-persistent SRS or CSI-RS

reference signals (SRS and CSI-RS) and to enable/disable CSI This option provides a trade-off between periodic and aperiodic
reporting. This allows the network operator to make a tradeoff where the network can quickly (without the need to use RRC)
between reference signal overhead (periodic vs aperiodic), activate or deactivate the configured periodic signal using the
triggering overhead etc. MAC CE which is scheduled to the UE using PDCCH+PDSCH.
In semi-persistent operation, the UE is configured a periodic
Periodic SRS or CSI-RS signal using RRC signaling and starts to receive (CSI-RS) or
This is typically used for shared downlink signals among multiple transmit (SRS) the signal only when MAC CE has activated (or
UE. The UE is configured a periodic signal using RRC signaling deactivated).
and continues to receive (CSI-RS) or transmit (SRS) until RRC
deactivates the signal. When it comes to CSI reporting, there are three possibilities to use
time dynamics:
Aperiodic SRS or CSI-RS
This is used for signals for one UE, e.g. when the network would • Periodic CSI reporting is configured by RRC and can only use
like a UE to perform a CSI measurement or get a report or the periodic CSI-RS for its measurements
network would like to perform the measurement on SRS for one
UE. • Semi-persistent CSI reporting is configured by RRC and
triggered by MAC CE (for PUCCH reporting) or DCI (for PUSCH
The UE is configured an aperiodic signal using RRC signaling reporting). It can use either periodic or semi-persistent CSI-RS
and DCI (i.e. PDCCH) is used to trigger the RS transmission/ for the measurements.
measurement. The benefit is low RS overhead as it is only
• Aperiodic CSI reporting is configured by RRC and triggered
transmitted when network needs information, but on the other
by DCI. The trigger information (states) can be reconfigured
hand it consumes PDCCH overhead as each UE must be triggered
using MAC CE. The aperiodic CSI report can use either periodic,
with an individual PDCCH. Medium access control channel
semi-persistent or aperiodic CSI-RS for the measurements. If
element (MAC CE) can be used to change the configuration of
aperiodic CSI-RS is used, then the CSI-RS can be in the same
aperiodic triggers. If aperiodic SRS is triggered, then the SRS
slot as the DCI that triggers the CSI report.
transmission can be in the same slot as the DCI that carried the
trigger (at least in FDD or in special slots in TDD).

LTE has 10 MIMO transmission modes (TM)

No chan
Abbrevia
3GPP TM UE synch- UE measurement UE demodulation

Release ronization RS RS for CSI report RS
8 1 CRS CRS CRS (1 layer)
Subcarrier
8 2 CRS CRS CRS (1 layer, Tx diversity)
8 3 CRS CRS CRS (1-4 layers)
8 4 CRS CRS CRS (1-4 layers) Control
CRS
8 5 CRS CRS CRS (1-4 layers) Data
0 1 2 3 4 5 6 7 8 9 10 11 12 13
8 6 CRS CRS CRS (1 layer)
Symbol
8 7 CRS CRS DM-RS (1 layer)
9 8 CRS CRS DM-RS (1-2 layers)
The CRS is always
10 9 CRS 8 port CSI-RS DM-RS (1-8 layers)
transmitted by the eNB
11 10 CRS + CSI-RS 8 port CSI-RS + CSI-IM DM-RS (1-8 layers)
13 10 CRS + CSI-RS 16 port CSI-RS + CSI-IM DM-RS (1-8 layers) UE synchronization
14 10 CRS + CSI-RS 32 port CSI-RS + CSI-IM DM-RS (1-8 layers)
is using CRS in all TMs
CRS: Cell specific reference signal | DM-RS: Demodulation reference signal | CSI: Channel state information | CSI-RS: CSI reference signal | CSI-IM: CSI interference measurement
LTE is more rigid than NR, e.g. • TM 4 uses a UE recommended precoder from a codebook of
pre-defined matrices
• Fixed subcarrier spacing 15 kHz
• TM 6 is a special configuration of TM 4 where only a single
• Fixed PDCCH position in slot layer is allowed to be scheduled. Hence rank is restricted to
• Cell-specific reference signals (CRS) are always transmitted one
in every slot
• TM 5 can be seen as a modification of TM 6 intended for
• 1, 2 or 4 CRS antenna ports depending on eNB configuration MU-MIMO
• CRS is an “all purpose” reference signal used for syn- • TM 7-8 use CRS for synchronization and measurements
chronization, measurement and demodulation (depending and DMRS for demodulation and was introduced to support
on transmission mode (TM) non-codebook based operation
LTE MIMO is defined by 10 transmission modes (TM) • TM 7 is a single PDSCH layer transmission can be used for UE
specific beamforming and reciprocity operation
• TM 1-6 use CRS for synchronization, measurement and
demodulation • TM 8 is a dual layer PDSCH layer transmission (Release 9)
• TM 1 and 2 are single PDSCH layer schemes (antenna • TM 9-10 use CRS for synchronization, CSI-RS for
diversity using 2 or 4 base station antennas) measurements and DMRS for demodulation to support non-
codebook based operation as well as multi-TRP transmission
• TM 3 and 4 are multi-layer PDSCH schemes (spectral and UE specific measurement RS (i.e. the CSI-RS)
efficiency up to 4 layers)
• TM 9 supports up to 8 layer PDSCH through the use of
• TM 3 uses a pre-defined precoding cycling through a set DM-RS defined for 8 ports (CRS is limited to 4 ports and is cell
of matrices from a codebook of matrices for a given rank. specific)
Obtains diversity for 2 or 4 eNB antennas.
• TM 10 supports up to 8 layer PDSCH plus multi-TRP operation
through enhanced CSI measurement and reporting

3GPP bands for NR FR1

N
NR operating Uplink (UL) operating band Downlink (DL) operating band NR operating
Duplex mode Uplink (UL) operating band Downlink (DL) operating band Duplex mode
band band
n1 1920 MHz – 1980 MHz 2110 MHz – 2170 MHz FDD n74 1427 MHz – 1470 MHz 1475 MHz – 1518 MHz FDD
n2 1850 MHz – 1910 MHz 1930 MHz – 1990 MHz FDD n75 N/A 1432 MHz – 1517 MHz SDL
n3 1710 MHz – 1785 MHz 1805 MHz – 1880 MHz FDD n76 N/A 1427 MHz – 1432 MHz SDL
n5 824 MHz – 849 MHz 869 MHz – 894 MHz FDD n77 3300 MHz – 4200 MHz 3300 MHz – 4200 MHz TDD
n12 699 MHz – 716 MHz 729 MHz – 746 MHz FDD n80 1710 MHz – 1785 MHz N/A SUL
n13 777 MHz – 787 MHz 746 MHz – 756 MHz FDD
n81 880 MHz – 915 MHz N/A SUL
n82 832 MHz – 862 MHz N/A SUL
n247 1626.5 MHz – 1660.5 MHz 1525 MHz – 1559 MHz FDD n84 1920 MHz – 1980 MHz N/A SUL
n25 1850 MHz – 1915 MHz 1930 MHz – 1995 MHz FDD n85 698 MHz – 716 MHz 728 MHz – 746 MHz FDD
n29 N/A 717 MHz – 728 MHz SDL n90 2496 MHz – 2690 MHz 2496 MHz – 2690 MHz TDD
n30 2305 MHz – 2315 MHz 2350 MHz – 2360 MHz FDD n91 832 MHz – 862 MHz 1427 MHz – 1432 MHz FDD (NOTE 2)
n34 2010 MHz – 2025 MHz 2010 MHz – 2025 MHz TDD
n92 832 MHz – 862 MHz 1432 MHz – 1517 MHz FDD (NOTE 2)
n93 880 MHz – 915 MHz 1427 MHz – 1432 MHz FDD (NOTE 2)
n40 2300 MHz – 2400 MHz 2300 MHz – 2400 MHz TDD n94 880 MHz – 915 MHz 1432 MHz – 1517 MHz FDD (NOTE 2)
n41 2496 MHz – 2690 MHz 2496 MHz – 2690 MHz TDD n95 2010 MHz – 2025 MHz N/A SUL
n46 5150 MHz – 5925 MHz 5150 MHz – 5925 MHz TDD (NOTE 3) (NOTE 1)
n48 3550 MHz – 3700 MHz 3550 MHz – 3700 MHz TDD n96 5925 MHz – 7125 MHz 5925 MHz – 7125 MHz TDD (NOTE 3)
n50 1432 MHz – 1517 MHz 1432 MHz – 1517 MHz TDD (NOTE 4)
n51 1427 MHz – 1432 MHz 1427 MHz – 1432 MHz TDD n975 2300 MHz – 2400 MHz N/A SUL
n53 2483.5 MHz – 2495 MHz 2483.5 MHz – 2495 MHz TDD n985 1880 MHz – 1920 MHz N/A SUL
n996 1626.5 MHz -1660.5 MHz N/A SUL
n100 874.4 MHz – 880 MHz 919.4 MHz – 925 MHz FDD
n67 N/A 738 MHz – 758 MHz SDL
n1024 5925 MHz – 6425 MHz 5925 MHz – 6425 MHz TDD3
This table shows the RAN4 defined bands for NR in FR1 NOTE 5: The requirements for this band are applicable only
spectrum. Copied from 3GPP specifications TS 38.104 where no other NR or E-UTRA TDD operating band(s) are used
within the frequency range of this band in the same geographical
FDD – Frequency division duplex area. For scenarios where other NR or E-UTRA TDD operating
band(s) are used within the frequency range of this band in the
TDD – Time division duplex
same geographical area, special co-existence requirements may
SDL – Supplementary downlink apply that are not covered by the 3GPP specifications.
SUL – Supplementary Uplink NOTE 6: UL operation is restricted to 1627.5 – 1637.5 MHz and
1646.5 – 1656.5 MHz per FCC Order DA 20-48.
NOTE 7: DL operation is restricted to 1526-1536 MHz frequency

NOTE 1: This band is applicable in China only. range. UL operation is restricted to 1627.5 – 1637.5 MHz and
1646.5 – 1656.5 MHz per FCC Order 20-51
NOTE 2: Variable duplex operation does not enable dynamic
variable duplex configuration by the network, and is used such NOTE 8: [This band is applicable only in countries/regions
that DL and UL frequency ranges are supported independently in designating this band for IMT licensed operation in accordance
any valid frequency range for the band. with RCC Recommendation 1/21.]
NOTE 3: This band is restricted to operation with shared

spectrum channel access as defined in TS 37.213.
NOTE 4: This band is applicable only in countries/regions

designating this band for shared-spectrum access use subject to
country-specific conditions.

SRS carrier switching
In some cases, the first n symbol

cannot be used to transfer data due
to carrier switching
Normal DL subframe
PCell 0 1 2 3 4 5 6 7
Special subframe
Normal UL subframe
Switch back to PCell carrier
Switch to SCell carrier Not available UL subframe
(due to switching or locked
to another CC)
SCell 0 1 2 3 4 5 6 7
SRS sent in last two

symbols
Typically, in carrier aggregation, the number of DL carriers The SRS carrier switching is an optional UE feature that can be
is larger than the number of UL carriers and fast SRS carrier configured to be periodic or triggered aperiodically. Also, carrier
switching is therefore supported in LTE and NR in order to switching can be combined with the antenna switching feature
transmit SRS on any UL carrier (even though such a carrier does if the UE also supports that. In addition, the switching time is a
not have possibility for PUSCH/PUCCH transmission). The UE UE capability that can range from 0 us to 900 us.
can therefore switch radio frequency (RF) chain (hardware
for signals reception/transmission) to SCell and temporarily
suspend the UL transmission on PCell. This is so-called SRS
carrier-based switching and is associated with an interruption
time in UL (and in some cases also DL) due to RF retuning, which
impacts throughput on the PCell (or PSCell for SCG based carrier
aggregation).

SRS enhancements
for Massive MIMO (Rel. 17) N
Max 4 SRS OFDM symbols. Anywhere within last 6 symbols of slot
SRS bandwidth
Release 15 Release 15
Time (one slot)
Example C Example B Example A

Release 17 Release 17
Up to 14 SRS OFDM symbols. Anywhere in the slot
To improve Massive MIMO with MU-MIMO scheduling with To improve Massive MIMO with MU-MIMO scheduling with
reciprocity-based operations, the aperiodic triggering procedure reciprocity-based operations, the coverage of SRS has been
for SRS has been enhanced (dynamic SRS trigger offset, left enhanced through transmission repetition (SRS capacity and
figure). coverage, right figure).
It is beneficial for MU-MIMO performance if the reciprocal Release 15 specification limitation:

channel can be measured at the same time instant for all An SRS resource only support 4 repetitions in a slot. Hence,
co-scheduled UEs. the coverage benefit (or interference reduction benefit if SRS
frequency hopping is enabled) is limited.
Release 15 specification limitation:
The time distance between the PDCCH that contains the SRS Release 17 enhancement:
trigger, and the actual transmission of the aperiodic SRS is Remove restrictions on SRS resource configuration. A Rel.17 UE
statically (RRC) configured. can be configured with up to 14 SRS OFDM symbols in a slot (i.e.
the whole slot) and the position of the SRS can also be anywhere
This leads to PDCCH congestion problem at high cell load since in the slot for maximal flexibility.
many UEs (which are candidates for MU-MIMO scheduling)
needs to be triggered from the same slot in order to have SRS Combined with frequency hopping is also supported.
transmitted in the same slot.
• Example A shows an 8 symbol SRS in the end of the slot
Release 17 enhancement:
Introduction of a slot offset indicator in PDCCH that allows for • Example B shows a 5 symbol SRS in the interior of the slot
dynamic indication of the time distance. • Example C shows a 14 symbol SRS spanning the whole slot
Thereby can the different PDCCH be distributed across multiple
slots and congestion is relieved.

NR provides several multi-TRP options
Rel. 16 NC-JT Single DCI Rel. 16 NC-JT Multi-DCI Rel. 17 HST-SFN Rel.18 CSI report for C-JT
Relaxed time sync (< CP) Relaxed time sync (< CP) Relaxed time sync (< CP) Relaxed time sync (< CP)
Relaxed frequency sync Relaxed frequency sync Relaxed frequency sync Tight frequency sync
Fast Backhaul Somewhat Relaxed Backhaul

Fast Backhaul Fast Backhaul
TRP1 TRP2 TRP1 TRP2 TRP1 TRP2 TRP1 TRP2
K2
DC -AC
PDSCH
1st set of layers PDSCH
Q
PD
I2
CS
AR
All layers
SC
I1
,H
H
,C
PDSCH
1
I2
DC
SI
CS
DC
PDSCH All layers

CS
DC SI,
2,
DC AR
CS
I1
C-
H2
I1
2nd set of layers

HA
I
,H
JT
I,
PDSCH
H
SC
AR
RQ
C
PDSCH All layers
PD
Q-
-A
Q-
HA
AC
CK
All layers
AC
RQ
K1
-A
CK
UE UE UE UE
In NR, multi-Transmission Point (TRP) operation has been In Rel.17, a new multi-TRP scheme was specified, intended for
standardized in Rel.16 and onwards. high speed train (HST) scenario using single frequency network
(SFN) transmission. This means that the network transmit two
In Rel.16, non-coherent joint transmission (NC-JT) was identical copies of a PDSCH from two TRPs simultaneously
introduced, where the UE receives from two TRP simultaneously and overlapping in the PRBs. This is different from NC-JT since
in the same physical resource blocks (PRBs). both TRPs transmit all layers. The SFN transmission is a well
know transparent technique to improve the SNR of the PDSCH
In single-DCI NC-JT, a single PDCCH schedules a single PDSCH
transmission (i.e. The UE need not know that SFN transmission is
with multiple MIMO layers. It is transmitted from two TRPs and
taking place) and is used in e.g mobile broadcast transmissions.
received by the UE, where the layers are divided into two groups,
The new aspect here is that it is not transparent to the UE,
one per TRP.
instead the UE is given two synchronization signals so it can
In multi-DCI NC-JT, two PDCCH (one per TRP), schedules one actually perform sync towards both TRPs simultaneously when
PDSCH each (one per TRP). Hence each PDSCH is transmitted receiving the SFN transmitted PDSCH.
from different TRPs and received simultaneously by the UE.
In Rel.18, the inter-cell multi-DCI NC-JT was further extended
In Rel.17, NC-JT was extended to support inter-cell operation, to include two time advance receive timing difference between
hence the two TRPs can belong to different cells (different the two DL reference timings (i.e. the two TRPS) and they can
physical cell ID). be assumed to be larger than cyclic prefix length. Hence, this
extends the use of multi-TRP to larger macro cells where the
In addition, a dedicated CSI report for single DCI NC-JT was receive timing difference can be large.
specified, which takes into account the fact that the UE is
receiving from two TRPs simultaneously. In Rel.18, CSI reporting for coherent joint transmissions (C-JT)
was introduced. In C-JT, similar to HST-SFN, two identical copies
of the PDSCH are transmitted from two TRPs simultaneously and
on the same PRBs. However, in this case, the transmissions are
phase aligned that gives coherent combining (constructive) of
the two PDSCH which boosts the SNR in the receiver. This type
of transmission requires tight frequency sync among the TRPs
and also CSI knowledge on how the network should perform the
co-phasing.

Beam management
Z
O
U
CSI-RS
PUSCH TRP
TRP TRP SRS
PDSCH SSB
PUSCH
UE
UE UE
QCL Type D (spatial RX parameter) for DL Spatial relation DL à UL Spatial relation UL à UL
Tx Rx Tx
Rx
UE UE
UE does not support UE supports

beam correspondence beam correspondence
Beam management is introduced in NR specifications, mainly A related concept for the UE side transmission is spatial
for the high band deployments. relation.
• It is needed when UE and/or gNB can only listen in one • If there is a spatial relation configured between a downlink
direction at a time signal and an uplink signal, such as PUSCH DM-RS and SSB,
then the UE transmit PUSCH in the same beam as it previously
• It requires an indication of “use same beam as before”, e.g. if received SSB
the gNB transmitted a first reference signal using a certain
beam, it can indicate to the UE that it will use the same beam to • This means that the gNB can this receive PUSCH in the same
transmit a second signal. Typically the first and second signals beam as it previously transmitted SSB
are a measurement signals (e.g. SSB) and a data or control
signals (e.g. PDCCH), respectively. This implies the UE can • This requires UE beam correspondence, meaning that the UE
safely use the same receive beam to receive the second signal support transmitting and receiving with the same beam. Not all
(provided that the UE has not rotated). UE support this and this is thus a UE capability.
• A key concept here is the Quasi co-location (QCL) framework, Moreover

since if a first and second reference signal is signaled to the
• If there is a spatial relation configured between two uplink
UE as QCL with respect to “Type-D (Spatial Rx parameter)”, it
signals, such as PUSCH DM-RS and SRS, then the UE transmit
implies that UE can use the same receive beam for these two
PUSCH in the same beam as it previously transmitted SRS
reference signals.
• This means that the gNB can this receive PUSCH in the same
• The UE typically need to keep track of multiple such QCL
beam as it previously received SRS
relations (i.e. multiple possible receive beams depending
of which transmit beam the network is using). Therefore, • This doesn’t require UE beam correspondence, since the UE
the 3GPP specification introduces a “TCI state” which a only use two occasions of a transmit beam
book-keeping framework for QCL relations and a UE can be
pre-configured with a large number of such TCI states, each
pointing to a downlink reference signal.

5. 3GPP—radio
requirements
5. 3GPP — radio requirements

Ericsson | The Massive MIMO Handbook — Technology Primer 5. 3GPP — radio requirements 63
Why radio requirements N
Base Station (BS) Radio Frequency (RF)

TM
Multi-standard
Single RAT AAS (Massive
radio
(5G only, MIMO)
(combinations of
4G only, base station
2G, 3G, 4G, 5G in
3G only) specification
same BS)
User Equipment Radio Base Station (BS) Electromagnetic Radio Resource

Frequency (UE RF) conformance compatibility (EMC) Management (RRM)
Spectrum is an expensive and scarce resource. Efficient Another important 3GPP RAN4 specification is the Radio
compatibility and co-existence in a multi-operator environment Resource Management (RRM) specification where UE behavior
is therefore essential. This is achieved by defining radio and performance in relation to measurements and procedures
requirements in 3GPP RAN4 and in regulatory specifications. The that support network operation are captured. The RRM
requirements consider co-existence between service providers specification has a significant impact on, amongst other things,
in mobile bands and in the case of regulation, between different network timing and synchronization.
types of systems. In addition to enabling efficient spectrum
usage, the RAN4 specifications also ensure consistent and Electro Magnetic Compatibility (EMC) is yet another RAN4
predictable behavior from UEs, which is essential for building specification covering both BS and UE that describes the
high performance and reliable networks. Separate specifications requirements for radiated emission and immunity aspects.
are defined for User Equipment (UE) and base stations (BS), The EMC specifications embed many of existing regulatory
above referred to as Massive MIMO radio and RRU. Note that specifications.
in 3GPP RAN4 and radio regulations, the term AAS is used
In addition to requirements, 3GPP also specifies the
for Massive MIMO base stations. In addition to spectrum
configurations, test set-ups and accuracy to be used for
requirements, the BS specifications contain other requirements
assessing compliance to the requirements. The BS test
such as transmitter signal quality or receiver behavior to provide
specification is a part of 3GPP RAN4 responsibility.
for good minimum network performance and coverage.
In summary, the 3GPP radio requirements provide a means to
The UE RF specification covers similar requirements as the BS
secure co-existence and compatibility, cover some regulatory
specification but is more extensive to ensure interoperability and
requirements and ensure good performance and predictability
minimum performance with different UE vendors. It contains
in respect to UE behavior in the network. In addition, the 3GPP
for example RF requirements for an extensive number of Carrier
RAN4 specifications provide a benchmarking tool for service
Aggregation (CA) and Dual Connectivity (DC) combinations
providers to enable fair comparison of performance between
because the hardware requirements for supporting different
different vendors under similar conditions.
combinations are different in the UE and in the BS. The
specification needs to contain suitable combinations for all
service providers needs.

Electro magnetic field (EMF) exposure

Electro magnetic field (EMF) exposure No chan
Wrong f
Strict limits on EMF

must be considered for
the whole site
All radio equipment including Massive MIMO (AAS) base A Massive MIMO (AAS) BS has the capability to do flexible
stations need to comply to regulation on electro-magnetic field beamforming both in azimuth and elevation, which is specifically
(EMF) exposure. EMF requirements are specified by international intended to obtain higher directional radiated power. The
bodies such as international commission on non-Ionizing assessment of EMF compliance for Massive MIMO (AAS) is
radiation protection (ICNIRP) or institute of electrical and more complex than for fixed antenna systems. The emissions
electronics engineers (IEEE), or national administrations such as depend on the type of traffic and the radiation pattern. The latter
federal communications commission (FCC). EMF is not, however, depends on the chosen multi-antenna transmission schemes.
in the scope of 3GPP. The EMF exposure limit is frequency
dependent and is expressed as incidence equivalent planewave With increased market demand for increased radiated power
power density (W/m2) averaged over time (6 minutes for from the Massive MIMO (AAS) BS, incorporating advanced
frequencies up to 10 GHz). algorithms to restrict the radiated power and to ensure
compliance to EMF exposure limits becomes a necessity.
The recommended EMF exposure limits provide a good margin
of protection against all known health effects. Some countries
require even stricter levels than the recommended ones.

Challenges for AAS conformance

requirements Z
Is
ar
Lo
Inte
gra ial
and t e d an pat
ne w te m ic s r
tech nna/ra a
Dy n e hav
io
nolo di b
gies o
New AAS
(Massive MIMO)
radio requirement
paradigm
No antenna
connectors
For Massive MIMO (AAS) there are new and additional When it comes to meeting requirements, AAS presents new
challenges in defining suitable requirements, designing and challenges. AAS arrays consist of a large number of individual
building products to meet the requirements, and in measuring low power transmitters, and the performance and efficiency
compliance. achieved with a single high power transmitter do not easily scale.
In addition, the very limited space in a tightly integrated array
Three key aspects of AAS impact the definition of requirements. puts extreme constraints on the possibilities for e.g. filtering,
The first is that the antenna elements are integrated into the which is key to meeting unwanted emissions requirements.
base station and antenna performance is part of the overall radio These aspects and a number of other technology considerations
performance. This differs from conventional base stations, for are explored on the next page.
which the antenna was a separate component independent from
the radio operation such that requirements could be set on the Achieving all radio requirements, including unwanted emissions
radio only. The second is that due to dynamic beamforming, the and receiver selectivity, drives the complexity and cost of AAS
spatial behavior of the wanted signal, unwanted emissions and solutions. In some cases, beamforming physics are able to
receiver is complex and time varying. For conventional (non- compensate radio RF performance for the wanted signal itself.
AAS) base stations, the spatial behavior is both static and quite However, beamforming cannot compensate radio performance
similar between different base stations. The third is that direct for unwanted emissions and receiver effects in the same way.
connectors to the radio output may not be available for AAS.
This is because efficient antenna array designs are not possible To capture the whole effect of the antenna array including
whilst providing connectors to each individual radio. The lack spatial properties and to enable assessment of the radio
of connectors makes traditional testing methods, in which test performance for AAS base stations that do not have connectors,
equipment is connected directly to the radio output, infeasible. over-the-air (OTA) testing is needed. Prior to 5G, OTA testing
of base stations was an unknown paradigm and much time
and effort have been spent in 3GPP to define requirements in
such a manner that OTA testing is possible and to investigate
the practical possibilities for OTA testing. The NR specification
contains OTA requirements for all parameters and test tolerances
based on realistic expectations of what can be achieved in test
environments.

3GPP over-the-air specification scope

No chan
Illustrative examples of beam patterns of advanced Massive MIMO systems
The purpose of the 3GPP radio specifications is to set minimum Beamforming and MIMO performance depend on baseband,
requirements to enable radio and the channel statistics. Current OTA test chamber
technology does not enable re-creation of fading channel
• Efficient sharing of spectrum between service providers and statistics in a controlled environment. For these reasons,
services determination and measurement of appropriate metrics for
estimating the ”quality” of beamforming and MIMO performance
• Predictable UE behavior
is non-trivial and is a subject to further research.
• A minimum signal fidelity
Even if such metrics would exist, it is not likely that it would
Requirements are generally absolute and must be met by all be suitable to set minimum requirements. To do so would
implementations. indirectly standardize the type of MIMO operation and risk to
restrict innovation. A potential future aim could be to determine
The radio specifications do not set requirements relating to the a toolbox of industry wide standardized metric descriptions
”quality” of the beamforming achieved. whose implications are understood, but which are not subject
to minimum levels. From these descriptions a suitable subset
Conventional fixed passive antenna performance was assessed
could be selected and assessed depending on beamforming
by measuring well-known parameters such as sidelobe levels,
implementation.
beamwidth, ripple of gain within the beam, front to back ratio
and others. For an AAS performing dynamic beamforming,
the parameters used for conventional BSs have less meaning
as they vary depending on the instantaneous beamforming.
Furthermore, for advanced MIMO systems the concept of a
”beam” is not applicable, see [TP3, p. 27]. Energy is radiated
in a complex pattern according to the channel state, as shown
in the figures. In the figures, the multiple lobes of energy are
deliberately created due to multi-path propagation and it is
not possible to differentiate ”useful” and ”not useful” radiated
energy.

What are the radio requirements

No
Transmitter requirements:
• Performance (power, signal quality,
Receiver demodulation requirements:
timing, ...)
• Link performance (physical channels
• Unwanted emissions (linearity,
under different fading conditions)
co-existence, ...)
• Co-location
Receiver requirements:
• Performance (sensitivity,
selectivity, linearity)
• Co-location
BS radio requirements consist in general of three main areas. The second area covers receiver aspects including receiver
sensitivity. Receiver sensitivity is defined as a minimum level
The first area is the transmitter requirements covering transmit of received signal that can be demodulated properly. Receiver
power related requirements, unwanted emission requirements sensitivity relates to the receiver noise figure, which is directly
and signal quality related requirements. The power related related to uplink coverage.
requirements cover the accuracy of the BS power, the transmitter
dynamic range requirements etc. Unwanted emission Apart from the receiver sensitivity, requirements are defined
requirements comprise (a) in-band requirements on absolute on receiver selectivity and blocking. The receiver blocking is a
operating band unwanted emission masks and adjacent measure of receiver´s capability to receive a weak wanted signal
channel leakage ratio (ACLR), which is a power ratio, regulating in the presence of a strong interferer in the adjacent channels.
emissions towards other service providers in the same 3GPP The receiver blocking thus defines the needed receiver in-band
band; and (b) out-of-band requirements comprising general selectivity. Receiver blocking can arise in the real world when
spurious emission requirements (in general set by regulators in another transmitter, such as another service provider’s UE is
different regions), which regulate co-existence to other types of close to the base station. In addition, there are co-location
system and co-existence and co-location requirements. Transmit blocking requirements specified considering very high interferer
signal quality requirements consist of error vector magnitude level in other bands.
(EVM), frequency error and time alignment error (TAE). EVM is a
measure of all transmitter impairments that degrade the signal The third requirements area covers receiver demodulation
quality. The frequency error captures the frequency variation performance. Receiver demodulation performance requirements
in relation to assigned frequency and TAE captures the time capture the performance of receiver baseband algorithms for
difference between different MIMO layers, or carriers when using different physical channels.
carrier aggregation (CA).

Classes of over-the-air requirement
Directional: Metric
measured in test direction
Test antenna
Total radiated power (TRP): Co-location: Power to/from a co-location

Sum of power radiated in all directions test antenna mounted next to the AAS
To define over-the-air (OTA) requirements with the same scope • The location of victims of emissions are not correlated
as the conducted requirements, it is useful to differentiate to the beamforming pattern
requirements based on the wanted signal (such as TX power and
dynamics, TX signal quality, RX sensitivity) and requirements • The unwanted emissions are often not beamformed
relating to other frequencies. • The impact of unwanted emissions to other systems is
often statistical in nature (such that average interference
For the wanted signal, there is always a single direction
relates to average impact towards other systems).
of interest, which is towards the intended UE.
For emissions and RX requirements related to co-located
For other frequencies, requirements are further differentiated equipment, requirements are defined based on a co-location
into those related to impacts on other radio receivers that are concept. A test antenna is placed 10 cm from the BS under
not at the same site and those related to avoiding interference test such that it experiences or causes interference in a similar
towards co-located equipment. manner to co-located equipment.
Most unwanted emissions requirements are specified based Some receiver requirements include both wanted signals and
on the so-called total radiated power (TRP) metric, which is interferers, such as blocker interferers on other carriers. These
the average power radiated in all directions. The reasons for requirements are also defined as directional with the interferer
selecting TRP are that: directions being the same as the wanted signal directions.
Although aligning the wanted signal and interferer directions
does not exactly correspond to real-life situations, the interferer
level is determined on a statistical basis such that the impact
to the radio electronics is similar to real life. Aligning the signals
enables a significant reduction in test complexity compared to
a set-up with signals coming from different directions.
The scope for spatial testing of baseband performance

requirements is limited; a test setup is defined in 3GPP but is
capable of testing at most polarization diversity.

Over-the-air testing
Over-the-air testing No
Examples of over-the-air
test facilities, depicting absorbant
material and an example of
mounting AAS (Massive MIMO)
base stations for testing
Measurement of OTA requirements with a reasonable degree of TRP requirements can be tested by means of making directional
accuracy is complex and specialized. A basic assumption is that measurements in many directions around the AAS such that
measurements are carried out in an RF shielded environment an average interference power can be calculated. Design of the
such that testing does not interfere with other nearby systems. measurement grid needs to be considered carefully and depends
on the geometry of the BS and the expected beam patterns. If
Conceptually, testing of directional requirements is not difficult; the test grid is not properly designed, the measurement results
the BS is illuminated from a particular direction or receives from may contain systematic and/or random error factors.
a particular direction. Design of testing is however complex:
An alternative to testing TRP by measuring in multiple directions
• The test facility needs to be designed such that a so-called is the use of a reverberation chamber (RC). A reverberation
”quiet zone”, in which only planar waves occur exists around chamber moves metal sheets within the chamber to randomize
the BS under test. a reflection pattern and enable average TRP to be measured. In
this case, care is needed to take care of calibration and of passive
• Careful calibration is needed to manage measurement
intermodulation effects.
uncertainties.
• The mechanical design of the test setup must also be Co-location testing requires the BS to be mounted together
sufficiently exact and carefully designed such that the AAS with the co-location test antenna. For some types of co-location
can be mechanically rotated, in order that the measurements requirements (in particular, transmitter intermodulation) it is
can be made from multiple directions, without leading to necessary to rotate the combination of BS and test antenna
degradation of accuracy. It should be noted that as long as the such that measurements of emissions can be made in multiple
BS can be rotated, indoor anechoic chamber (IAC), compact directions. For most types of co-location measurements, the
antenna test range (CATR), planar wave chamber (PWC) measurement is made in the co-location test antenna or in the
and near-field test chamber (NFTC) types of chambers are all BS, and the only requirement for the test chamber is that it is
suitable. shielded and anechoic.

Types of over-the-air testing chamber (I)

Types of over-the-air testing chamber (I) No chan
Wrong f
Example schematic diagrams
Test antenna Planar test

AAS Adaptive or AAS RF mirror AAS
antenna
θ θ
fixed array θ
φ φ φ
0 combining / 0 Test antenna 0
splitting
Test equipment
Indoor Anechoic Planar Wave Compact Antenna Test

Chamber (IAC) Chamber (PWC) Range (CATR)
There are several kinds of OTA test facilities that are relevant for Planar wave chamber (PWC): In a PWC, a steerable array is
base station testing. Not all of the facilities are suitable for every used to transmit to / measure from the AAS. The beamforming
type of requirement. For testing the full set of requirements, it is weights at the array are set such that the AAS experiences a
likely that more than one type of test facility is needed. planar wave as if from a source in the far field. The PWC is similar
in concept to a CATR, except that the RF mirror is replaced by the
Indoor anechoic chamber (IAC): The indoor anechoic chamber electronic array and is much smaller in size than an IAC.
is a shielded chamber with absorption such that a quiet zone is
achieved around the AAS. The AAS is placed in the far field of the IAC, CATR and PWC are generally suitable for all types of
test antenna. IACs operate within a range of frequencies they are in-band measurements and some out of band measurements
defined for. A disadvantage with an IAC is that it may be large in that are close to the band in frequency.
size.
Compact antenna test range (CATR): A CATR is a shielded,

anechoic chamber with a specially designed reflector (RF mirror)
between the AAS and the test antenna. A planar wave and quiet
zone is achieved at the AAS under test, and the dimensions of the
chamber can be reduced significantly.

Types of over-the-air testing

chamber (II)
Types of over-the-air testing chamber (II) N
Example schematic diagrams
Reflective
Moving walls
metal Near field
Test antenna
reflectors probes
AAS
AAS θ
Test equipment
θ φ
0
φ
0
Reverberation Chamber (RC) Near Field Test Chamber (NFTC)
Near Field Test Chamber (NFTC): A near field test facility is a type Electromagnetic Compatibility (EMC) chamber: EMC chambers
of chamber in which test probes are placed within the near field are shielded and somewhat anechoic and cover a large
of the AAS under test. The size is considerably reduced. NFTC frequency range. The uncertainties associated with EMC
is suitable for assessment of in-band TRP and some of the TX chamber measurements are larger than for other chambers.
directional requirements, but not RX requirements. EMC chambers are useful for measuring out of band emissions
because they cover a wide frequency range and the increased
Reverberation Chamber (RC): A reverberation chamber is uncertainty does often not matter because the emissions far from
shielded, but deliberately not anechoic. The inside walls are the carrier frequency passes the requirement with a large margin
metallic, and metallic moving ”stirrers” cause the profile of for most BSs.
reflections to continuously shift. The shifting reflection patterns
mean that it is possible to measure total radiated power
(TRP). RCs are suitable for measuring TRP based emissions
requirements for in band and parts of out of band, but not the
entire out of band range.

Summary
No chang
• 3GPP and regulatory radio requirements enable • It is critical to be able to validate that a Massive
efficient usage of the spectrum with predictable MIMO/AAS BS meets all relevant requirements
performance and behavior in the networks
• Proper OTA testing is specialized and costly.
• Accommodating Massive MIMO/AAS BS However, if OTA testing is not done properly,
necessitated completely new approaches to the Massive MIMO/AAS BS behavior and
define and measure the radio and baseband performance may be poor which can endanger
requirements over the air the eco-system
• Designing a Massive MIMO/AAS BS to meet the
over-the-air (OTA) requirements is complex and
challenging

Ericsson | The Massive MIMO Handbook — Technology Primer 8. mmWave technology 73
8. mmWave technology
6. Architecture
and implementation
7. Massive MIMO features

Ericsson | The Massive MIMO Handbook — Technology Primer 6. Architecture and implementation 75
Introduction
Introduction T
F
N
The commercialization of Massive MIMO radio units

has led to some key architecture choices in order to
realize cost-efficient hardware implementations
Fronthaul evolution
• Enhanced common public radio (eCPRI) interface
• The O-RAN Alliance
• Lower layer split
Beamforming
• Analog versus digital beamforming
• Frequency-domain versus time-domain beamforming
The commercialization of Massive MIMO radio units has led to Beamforming, which is the function that maps information
some key architecture choices in order to realize cost-efficient signals on multiple antennas, can be implemented in different
hardware implementations. ways. One choice is if it should be implemented with analog or
digital components. For the choice of digital beamforming, there
To support a large number of radio chains, the traditional is also a choice between frequency-domain or time-domain
common public radio interface (CPRI) has evolved in the beamforming.
enhanced common public radio interface (eCPRI), which
supports the so-called lower layer split, where some of the
traditional baseband functions have been moved to the radio
unit. Depending on which functions that are moved, different
lower layer splits can be defined.
6. Architecture and implementation

Hardware architecture
Hardware architecture
Baseband unit Massive MIMO radios

• Massive MIMO radios have more eCPRI
Digital Analog
radio chains than conventional Layer 2 Layer 1
front-end front-end
radios. To handle this in a cost-
BB ASIC (with DSP), CPU Radio ASIC Analog HW
efficient way, there are hardware
impacts/choices on the system Layer 2 Layer 1
Beam-
Beamforming
forming
architecture
Layer 1
• The location of the beamforming
function
Analog front-end processing
• The choice of digital beamforming Digital front-end processing CPRI or
Remote radios
Beamforming processing eCPRI
or analog beamforming Digital Analog front-
Layer 1 processing
front-end end
Layer 2 processing
Hardware platform Radio ASIC Analog HW
Massive MIMO radios have typically many more radio chains Beamforming can be implemented either using digital
than conventional radios, e.g. 64 radio chains instead of 4 radio components or analog components, or a combination of these.
chains. The traditional common public radio interface (CPRI) Typically, digital beamforming provides higher flexibility and
fronthaul protocol transfers data per radio chain. For Massive higher performance. However, for cases with very many radio
MIMO radios this would mean very large bitrates over the chains and/or very large frequency bandwidth, such as for
fronthaul interface. To handle this in a cost-efficient way, there mmWave products, the complexity and cost can become large,
are hardware impacts/choices on the system architecture. and in those cases analog beamforming is an alternative.
In order to lower these bitrates, the beamforming function

has been moved from the RAN Compute (e.g baseband) to
the radio. This makes the fronthaul bitrate scale with the user
information bitrate, rather than with the number of radio chains.
And since the user information bitrate typically is much lower
than the combined bitrates for all radio chains, this makes a large
hardware (HW) saving.

Fronthaul interface standardization N
Common public radio interface (CPRI)

• Proven interface for classic radios with few antennas
• Bit rates scales with the number of radio chains
Enhanced common public radio interface (eCPRI)
• Ethernet packet-based
• Enables new functional splits between baseband and
radios that can reduce the bit rate over the fronthaul
for Massive MIMO radio units
Standardization of new functional splits
• Studied in 3GPP
• Standardized by the O-RAN Alliance
The common public radio interface (CPRI) and enhanced CPRI The CPRI Forum did not standardize the functional split between
(eCPRI) are standardized by the CPRI Forum. the baseband and the radio. 3GPP did a study on different
functional splits but decided to not standardize which split
CPRI is a proven interface for Classic radios with few radio to use. The O-RAN Alliance started in 2018 with the aim to
chains. The connection is point-to-point, as a circuit switched ”Leading the industry towards open, interoperable interfaces and
network, and bitrates scale with the number of radio chains. RAN virtualization”. To standardize functional splits between the
baseband and the radios is the task of one of the working groups
The purpose of creating a new fronthaul interface, eCPRI, was
in O-RAN.
to enable more efficient implementations of Massive MIMO
radios with many radio chains, and to introduce Ethernet based As of today (2022), the O-RAN Alliance has standardized two
transport. An Ethernet based transport enables switched different categories of the so-called lower layer split (LLS), and
front-haul. several extensions and options.

Lower layer split (LLS) options

Minor ch
Note:
Should b
Baseband unit
• 3GPP option 6: MAC – PHY split MAC MAC
Option 6
• E-LLS: Ericsson LLS split for Coding De-coding
mid-band Massive MIMO Rate Matching Rate Matching
• O-RAN split: Scrambling

De-scrambling
Option 7-3
De-modulation
• Category A: precoding in baseband unit Modulation
iDFT
• Category B: precoding in radio unit Layer Mapping E-LLS PHY
Option 7-2 Equalization
• Standardization work started 2022 to move Precoding TX Power O-RAN Cat-B
more uplink functionality to the radio unit. Channel Estimation
Beamforming
• 3GPP option 8: CPRI split Option 7-1 O-RAN Cat-A Beamforming
iFFT
• Ericsson LLS split for remote radios and for FFT
mmWave Massive MIMO radios is similar to Cyclic Prefix Insertion Cyclic Prefix Removal
Option 8
O-RAN Category A split
RF Radio unit RF
The so-called lower layer split (LLS), is a functional split between The O-RAN Alliance has standardized two variants of the LLS.
a baseband unit and a radio unit. The 3GPP study (3GPP TR The Category A and Category B splits, which are similar to
38.816) on functional splits included the “Option”-alternatives the 3GPP Option 7-2. Category A is typically used for remote
shown in the picture. radio units, and Category B for Massive MIMO radios. Since
the beginning of 2022, O-RAN work group 4 has been running
Option 6 is a split between the media access control (MAC) layer a work item to study uplink performance improvements for
and the physical (PHY) layer. The payload information conveyed Massive-MIMO. The purpose of the work item is to be able to
over that split consists of transport blocks per layer. prove the same performance for an O-RAN compatible radio unit
as for an E-LLS radio unit.
Options 7-1, 7-2 and 7-3 are different splits within the PHY layer.
The Ericsson LLS split for remote radio units and for mmWave
Option 8 is a split between the PHY layer and radio processing
Massive MIMO radios is similar to O-RAN Category A split.
(which can include both digital and analog processing). The
payload information conveyed over that split consists of time-
domain IQ samples per radio chain.
Ericsson Massive MIMO radio units for mid-band uses the split
E-LLS, that for the downlink has precoding, beamforming, iFFT
and cyclic prefix insertion in the radio unit, and for the uplink has
cyclic prefix removal, FFT, beamforming, channel estimation and
equalization in the radio unit. The payload information conveyed
over this split consists of modulated bits per layer in downlink
and equalized IQ samples per layer in uplink.

Performance difference
Mid-band Massive MIMO downlink
MAC Baseband unit

O-RAN lower layer split (LLS) versus E-LLS Option 6
Coding
• No time-critical control loops that impacts
Rate Matching
the basic throughput performance in the
downlink Option 7-3
Scrambling
• Therefore, no basic beamforming Modulation
performance difference between different Layer Mapping E-LLS

Option 7-2
splits
Precoding TX Power O-RAN Cat-B
• Differences in fronthaul bitrates
Beamforming O-RAN Cat-A
Option 7-1
iFFT
Cyclic Prefix Insertion

Option 8
RF Radio unit
For the downlink, there are no time-critical control loops that For example, both E-LLS and O-RAN Cat B support efficient
impact the basic throughput performance, and the user data so-called “modulation compression” of the user data (not shown
functionality is the same for all splits. Therefore, the user in the picture), which O-RAN Cat A does not support. Therefore
throughput performance is basically the same for the different E-LLS and O-RAN Cat B will have lower fronthaul bitrates
splits. compared with O-RAN Cat A.
However, the different flavors of splits can give different front-

haul bitrates of the user data, and they will also have different
control signaling, which indirectly could lead to differences in
user throughput.

Performance difference
Mid-band Massive MIMO uplink
MAC Baseband unit

O-RAN lower layer split (LLS) vs. E-LLS
De-coding
• Time-critical control loop that affects the original Rate matching

Cat-A/Cat-B O-RAN splits De-scrambling
• It takes too long time to transfer the channel De -modulation

information back and forth between the radio
iDFT
and baseband units E-LLS
Equalization
• Therefore, the beamforming operation in the Information Closely
Channel estimation
radio unit will be based on channel estimates related
from previous slots Beamforming
O-RAN Cat-B
• A work item started in O-RAN in 2022 to enable FFT
same performance as E-LLS. Cyclic prefix removal
RF Radio unit
For the uplink, there are tight relations between some of the This means delayed information of at least several milliseconds,
functional blocks, the beamforming, channel estimation and even 10s of milliseconds, which gives performance issues
equalization. In particular, channel estimation has time-critical both for the case with high velocity UEs and for the case with
dependency on the beamforming. significant inter-cell interference.
This creates issues for the original O-RAN mid-band Massive This leads to a trade-off between user throughput and fronthaul
MIMO uplink. With channel estimation in the baseband unit, bit rates. By using excessive front haul bit rates, the performance
the beam weights applied in the radio unit will be based on old loss can be minimized. But on the other hand, if the same
channel estimates, due to the processing and transfer delay of fronthaul bit rates would be used for the E-LLS split and the
sending the information back and forth between the radio unit original O-RAN splits, there will be a significant user throughput
and the baseband unit. And even if the pure transmission delay loss for the original O-RAN splits.
is not that large, it is still too much to get the information back in
time to be able to apply the beam weights based on the received To avoid this trade-off, a work Item in O-RAN started in 2022 to
reference signals in the same slot. Instead, the beam weights enable same performance as E-LLS.
need to be calculated on received SRS or DMRS from previous
slots.

The Ericsson LLS advantage

N
• Ericsson is technical leader in lower layer split

Superior architecture
(LLS).
• The maturity of Ericsson Massive MIMO Full IRC receiver
products gives performance advantages.
Custom silicon
• There is a clear uplink performance advantage
of E-LLS over the original O-RAN LLS for mid- DL beamforming
band Massive MIMO uplink, which can either
translate to higher user throughput or less eCPRI
fronthaul fibers.
• A work item started in O-RAN in 2022 to enable
same performance as E-LLS.
Baseband unit
Ericsson is technical leader in lower-layer split (LLS). Ericsson The maturity of Massive MIMO products and their E-LLS
was the driver of the eCPRI specification in the CPRI Forum and interface give a performance advantage. And this is not only
was first with eCPRI based products for Massive MIMO. Several about the better uplink performance for Massive MIMO, but also
years before the O-RAN Alliance was created. that E-LLS has better support for various radio features, e.g.,
power saving features, passive intermodulation (PIM) detection/
Ericsson have worked on LLS for many years and have avoidance.
now ASICs that optimize performance, costs and power
consumptions. There is a trade-off between user throughput and fronthaul bit
rates for the current O-RAN Cat-A/Cat-B splits for the Massive
MIMO uplink. A work item started in O-RAN in 2022 to enable
same performance as E-LLS.

Open interfaces does not equal

multi-vendor
Open interfaces does not equal No chan
multi-vendor Change
Multi-vendor interoperability is the major While multi-vendor interoperability over the

driving force for the O-RAN LLS from a fronthaul could offer the benefit of an increased
service provider perspective: ecosystem, several challenges are also introduced:
System integration Life-cycle management
4 Fundamental
Challenges
System performance
Assurance of
KPIs and security
The major driving force for the O-RAN lower-layer split (LLS) System performance—Minimum common denominator
from a service provider perspective is the possibility to connect dictates feature support by the vendors involved, resulting in
baseband and radio units from different vendors, i.e. multi- performance limitations.
vendor interoperability.
Assurance of key performance indicators (KPIs) and security
However, open interfaces does not equal multi-vendor. Multi- —Challenging root cause analysis to identify vendor at fault
vendor introduces several challenges. and who is responsible for providing fixes.
System integration—Extensive integration projects to verify Example: The air interface is fully standardized in 3GPP,
an open interface between vendors add time-to-market (TTM) including detailed test specifications, but still requires extensive
and cost to the solution. interoperability device testing (IODT) between the RAN vendors
and the chipset vendors.
Life-cycle management—Software releases between vendors
need to be coordinated, tested and verified to ensure that
interoperability is maintained.

Analog versus digital and time domain

versus frequency
Analog versus digitaldomain
and time domain S
versus frequency domain
Signal processing iDFT/DFT Signal processing AD/DA Signal processing

Implementation
view
Digital beamforming Analog beamforming

Functional
view
Frequency-domain Time-domain
beamforming beamforming
DFT: discrete Fourier transform | iDFT: inverse discrete Fourier transform | AD: analog-to-digital (converter) | DA: digital-to-analog (converter)
This picture illustrates different beamforming types. The reception can be described in a similar way, but with a signal
flow starting from the right.
In the transmission direction, the signal flow starts from the left
with a signal processing block, which is then followed by an iDFT/ Digital beamforming = beamforming in the digital domain
DFT that translates the signal from the frequency domain to the
time domain. After another signal processing block, the signal is Analog beamforming = beamforming in the analog domain
transferred from the digital domain to the analog domain in the
Frequency-domain beamforming is always performed in
AD/DA block. Finally, some signal processing is performed before
the digital beamforming. Time-domain beamforming can be
the signal reaches the air.
performed by digital beamforming, analog beamforming, or a
combination of both digital and analog beamforming. Further
details follow on the next page.

Beamforming implementations
Beamforming implementations
Time domain beamforming versus frequency domain Time-domain

Frequency UE1 UE4
beamforming beamforming
(PRBs) UE2 UE5
One beamformer
• Time-domain beamforming few beamformers UE3 UE6
• Frequency-domain beamforming many beamformers
Digital beamforming versus analog beamforming

Time
Slot
• Digital beamforming enables frequency-domain
beamforming Frequency-domain
beamforming Frequency
• Total beamforming complexity: number of antennas Beamformers
(PRBs)
times the number of beamformers
• If the total complexity becomes too high, analog

beamforming, using time-domain beamforming, reduces
the complexity, but lowers the throughput performance
Slot Time
Beamforming can be implemented either in the frequency Frequency-domain beamforming requires digital beamforming.
domain or in the time domain. The fundamental difference Time domain beamforming can be implemented either by
between these two implementations is the frequency resolution digital beamforming or analog beamforming. If implemented
of the signal processing. Typically, frequency-domain in the digital domain, it can be done per carrier, while an analog
beamforming operates on a resource block level (typically a few implementation is normally done for the total bandwidth of the
100 kHz), and time domain beamforming typically operates on a radio.
carrier level (typically in the order of 100 MHz).
The total beamforming complexity depends both on the
A beamformer is an entity that operates on a frequency/time frequency resolution and the number of radio chains. If the total
block with one set of beamforming weights. Frequency-domain complexity becomes too large, using analog beamforming (and
beamforming has more beamformers than time-domain time-domain beamforming) is a way to reduce the complexity,
beamforming, due to its finer frequency granularity. This finer but at the expense of lower performance.
granularity gives a higher level of control, and therefore an
expected higher performance. The following implementations are typical choices of
beamforming:
• Mid-band 100 MHz bandwidth, 64 antennas Digital

frequency-domain beamforming
• High-band (mmWave) 800 MHz bandwidth, 512 antennas

Analog time domain beamforming

Architecture and radio chains

Architecture and radio chains
Digital beamforming architecture each polarization has one digital radio chain Digital radio chain Analog radio chain
and one analog radio chain per subarray
Digital processing Analog processing incl up-
incl radio signal conversion, PA and filtering
iDFT RSP DAC PA processing (RSP)
..
𝑊𝑊𝑑𝑑 .
up-conversion
iDFT RSP DAC PA
subarray
up-conversion
and phase adjustment
Analog beamforming architecture each polarization has one digital radio chain
iDFT inverse discrete Fourier transform
and multiple analog radio chains (one per subarray)
RSP radio signal processing:
crest factor reduction, digital predistortion, etc.
DAC digital-to-analog converter
iDFT RSP DAC
..
𝑊𝑊𝑑𝑑 . PA power amplifier
𝑊𝑊𝑑𝑑 digital frequency-domain beamforming
iDFT RSP DAC
! $ 𝑊𝑊𝑎𝑎 analog time-domain beamforming
As discussed previously, there are two main choices for In principle, it is possible to beamform each resource element
implementing beamforming: independently, which makes it possible to transmit multiple
differently beamformed layers using all the antennas on all
• A digital implementation, where the beamformer typically resources and to do generalized beamforming where a layer can
operates on frequency-domain complex-valued baseband be differently beamformed on different frequencies to match
signals the channel conditions in both space and frequency domains.
Each of the 2NM frequency-domain signals are then individually
• An analog implementation, where the beamformer typically
converted to time-domain using an iDFT and then mapped to
operates on time-domain real-valued radio frequency signals
a digital radio chain comprising filtering, crest factor reduction
For a digital baseband implementation, the beamforming (CFR), digital pre-distortion (DPD), etc. Each digital radio chain
function is implemented in the frequency domain before the is then converted to analog domain using a digital-to-analog
inverse discrete Fourier transform (iDFT). For a dual-polarized converter (DAC) and mapped to a corresponding analog radio
NxM array of subarrays, the beamformer generates a chain. The analog radio chain comprises up-conversion to radio
2NM-dimensional signal for each subcarrier in each OFDM frequency, power amplification with a PA and analog filtering,
symbol, i.e. for each resource element. thus converting a low-power baseband signal to a ‘high-power’
radio frequency signal. Finally, each of the analog radio
chains are mapped to an associated subarray. Thus, for digital
beamforming, there is one digital radio chain and one associated
analog radio chain for each of the signals generated by the
digital beamformer. The total number of radio chains (digital
and analog) equals two times the number of subarrays for a
dual-polarized array of subarrays.

Architecture and radio chains cont.
For analog beamforming, the beamforming is performed in the For both digital and analog beamforming there are 2NM analog
analog domain. This means that an analog radio chain in this radio chains for a dual-polarized NxM array of subarrays
case also comprises means to adjust the phase and amplitude of configuration. For digital beamforming, there is a one-to-one
the radio frequency signal. An analog beamformer operates on mapping between digital and analog radio chains, hence
a set of analog radio chains, and a key difference as compared 2NM digital radio chains for digital beamforming. For analog
to digital beamforming is that there is (typically) only one digital beamforming there is only 2 digital radio chains, one per
radio chain per analog beamformer. That is, a single digital polarization (one per analog beamformer). Although the analog
radio chain is, after the DAC, fed to an analog beamformer radio chain is slightly more complex for the analog beamforming
block that consists of several analog radio chains that can implementation due to e.g. gain and phase adjustments, the
apply individual gain and phase adjustments according to the main difference in complexity between a digital and an analog
beamforming objective. The signal from each of the analog architecture comes from the number of digital radio chains and
radio chains in a beamforming block is then mapped to one of associated interface requirement differences.
the two polarizations of a dual-polarized subarray. Hence, for a
dual-polarized NxM array of subarrays, each analog beamformer Note that the architecture descriptions presented here are
consists of NM analog radio chains, and as the array consists indicative, but exact implementations can vary. Also, it is possible
of two polarizations, two analog beamformers are used, one to mix a digital and an analog beamforming architecture, where
per polarization. Hence, in total two digital radio chains and parts of the beamforming is done in the digital domain and parts
2NM analog radio chains. Note that an analog beamforming in the analog domain. This is referred to as hybrid beamforming
architecture typically includes some degree of digital and is explained further on the following page. Different
beamforming as well. For example, for the analog architecture solutions can be envisioned, and strict definitions are difficult to
described here, digital beamforming is used to map one (or two) make.
layers over the two polarizations.

Hybrid beamforming
Hybrid beamforming F
The term hybrid beamforming UE1 UE4

is used if both frequency-domain UE1
UE2 UE5
and time-domain beamforming View from above
Vertical beam #1 UE2
are used UE3 UE6
UE3
Vertical beam #2
UE4
View from above
UE5
Vertical beam #2
View UE6
from side Vertical beam #1 UE1/2/3 UE4/5/6
The term hybrid beamforming is normally used to describe One example of hybrid beamforming is illustrated in the slide:
a system where digital and analog beamforming are
combined. Hybrid beamforming can also be used to describe The time-domain beamformer is creating a horizontally wide
the combination of frequency domain and time domain and vertically narrow UE-common beam, which is tilted in
beamforming regardless if the time domain beamformer is different vertical directions in different time slots. This is
implemented in digital or analog domain. However, in the first illustrated in the lower left figure, where two different time-
case, the digital beamforming is usually assumed to be in the domain vertical beams are shown. In each time slot, one of these
frequency domain and the analog beamforming in the time two beams can be used. For example, when scheduling UE1-3,
domain. beam #1 is preferred and when scheduling UE4-6, beam #2 is
preferred. Once the time-domain beam has been chosen, the
Hybrid beamforming provides a solution that lies in-between frequency domain beamformer can create UE-specific beams
full frequency-domain beamforming and full time-domain within the time-domain beam. This is illustrated to the right,
beamforming, both in terms of performance flexibility and where different UEs are (frequency or spatial) multiplexed using
implementation cost. different beams. Different UEs are then scheduled in different
time slots depending on the location, and it is primarily the choice
of the time-domain beam that dictates the set of UEs that can be
scheduled.

Summary—Key points
Summary – Key points Frequenc
The development of Massive MIMO radios has Beamforming types:

led to two important architecture choices, the • Digital beamforming in the frequency-
fronthaul solution and the beamforming type domain is preferable for radios with limited
number of antennas and limited frequency
The enhanced common public radio interface bandwidth, typically for mid-band products
(eCPRI) fronthaul: • Analog beamforming in the time domain is
• Enables lower fronthaul bitrates by preferable for radio units with large number
supporting more efficient functional splits of antennas and large frequency bandwidth,
typically for mmWave products
• Enables switched fronthaul solutions due
to its Ethernet-based transport • Hybrid beamforming leverages on the
strengths of both digital and analog
beamforming
The development of Massive MIMO radio units has led to two For high-band (mmWave) products, analog beamforming in the
important architecture choices, the fronthaul solution and the time domain is typically preferable as the radio units have both
beamforming type. many antennas and a large frequency bandwidth (compared
with mid-band products).
The eCPRI fronthaul interface enables lower fronthaul bitrates,
by supporting more efficient functional splits, and enables Hybrid beamforming provides a solution that lies in-between
switched fronthaul solutions due to its Ethernet-based transport. full frequency-domain beamforming and full time-domain
beamforming, both in terms of performance and implementation
For mid-band products, digital beamforming in the frequency cost.
domain is typically preferable as the radio units have both a
limited number of antennas and a limited frequency bandwidth
(compared with mmWave products).


7. Massive MIMO
features

Ericsson | The Massive MIMO Handbook — Technology Primer 7. Massive MIMO features 91
Introduction
Introduction No
Advanced MIMO for

Coverage
extreme performance
Ease-of-deployment, cost
and energy efficiency
The purpose of this chapter is to elaborate on the Massive MIMO Capacity and performance feature on the other hand typically
features introduced in [Ch. 3.3]. We can categorize the possible aims to improve peak user throughput and/or capacity of
Massive MIMO features in four different categories, which will be the data channels and generally uses a combination of
further elaborated on in the coming slides. beamforming, nullforming and spatial multiplexing. Advanced
MIMO features targets more specific deployments or scenarios
1. Coverage which may have an extreme capacity need, very high load
2. Capacity and Performance or some other special characteristics. For these cases, more
advanced e.g. inter-cell coordination features or higher-order
3. Advanced MIMO feature for extreme performance spatial multiplexing can be applied (which would not be required
4. Ease-of-deployment, cost and energy efficiency or provide additional benefit for e.g. less capacity-demanding
cells).
Coverage features intend to improve either the access coverage
or the data channel coverage typically by applying beamforming Features in the Deployability, Cost & Energy Efficiency category,
to boost the signal strength. This can be done either in uplink or on the other hand addresses non-performance related key
downlink. performance indicators (KPI) [Ch. 2, p. 36] such as e.g. fulfilling
regulatory requirements on emissions (EMF) which may
require dynamically adapting the power level depending on the
beamforming characteristics of the cell or reducing the amount
of beamforming that is performed in order to conserve energy.

7. 1 Coverage

Types of coverage improvements

Types of coverage improvements N
Access improvement Data channel improvement General coverage improvement

Feature which increases the link Improves both access and data
Feature which improves the
budget of a limiting control channel so channel coverage
coverage/performance of the traffic
that the UE can connect to or maintain
channel (”App coverage”)
a connection to the network in poorer
coverage
User performance User performance User performance
User Throughput
Users Throughput
User Throughput
SNR SNR SNR
When addressing coverage enhancements, it is important A data channel coverage improvement on the other hand will
to differentiate between access coverage and data channel manifest itself as a shift of the entire link curve to the left as is
coverage. illustrated in the middle picture, meaning that a certain user
throughput can be achieved at a lower SNR after the coverage
Access coverage can further be broken down into initial access improvement. This is also commonly referred to as improving the
coverage, which is the coverage of the cell-defining signals and app coverage, considering that a certain application may have
messages in the initial access procedure. All of which needs to be a requirement on X Mbps user throughput to properly function.
in coverage for a UE to find and connect to the cell, and control After the data channel coverage improvement we can achieve
channel coverage, which is the coverage of the supporting this at a lower SNR level and thus in a larger part of the cell.
control channels needed for scheduling and providing feedback
of/for the data channels. For example, an access coverage Then there are of course general coverage enhancements which
limitation manifests itself as a sharp drop to low user throughput can improve both access and data channel coverage
when the SNR reaches a low enough value, as is illustrated on (e.g. increasing the transmission power).
the red curve of the left picture. The interpretation is that the
data channel could be sustained and provide service to the UE at
even lower SNR levels (with smoothly and gradually decreasing
user throughput), but one or more of the control channels are out
of coverage and therefore not able to support the data channel.
By improving the coverage of said control channels so that they
are no longer limiting data channel performance, it is possible to
reach the performance of the blue curve.

Improving access coverage

Improving access coverage
Coverage can be defined as the

ability to serve traffic at a certain
rate
Control Control DL Data Before coverage
● All procedures and physical channel 1 channel 2 channel improvement
channels required to serve
the traffic need to have Effective cell range
coverage,
data channels, control Coverage [dB]
messages, initial access
messages Control Control Control DL Data After coverage
channel 1 channel 2 channel 1 channel improvement
Different messages have
different baseline link budget
New effective cell range
Link budgets can be different
in different implementations Control channel 1 Coverage [dB]
improvement
The coverage of the supporting channels needs to be on the PUSCH:

same level as the downlink/uplink data channel, otherwise
there is a coverage limitation. However, larger “coverage” of • Initial access coverage
the supporting channels does not necessarily result in actual
• Scheduling request (SR) using PUCCH
improved downlink or uplink coverage. Ideally all supporting
channels of a traffic channel should have the same coverage as • Grant carried on PDCCH
the traffic channel itself. If any supporting channel has smaller
coverage, that channel will be limiting and the traffic channel is • Data on PUSCH
not operational, but if a supporting channel has larger coverage
than the traffic channels it supports it provide no additional
benefit. That is, it is over-dimensioned which may have a Initial access:
detrimental impact on system performance as the improving the
coverage of a channel generally implies that more overhead is • Synchronization and SIB1
required (which means less resources for the data channel itself).
• PSS/SSS/PBCH
To support the data channels, a number of supporting channels
• PRACH (MSG 1)
needs to be in coverage as well, see [TP4, p. 46] for further
details. • MSG2 (RAR) carried on PDSCH with associated PDCCH
PDSCH: • MSG 3 carried on PUSCH
• Initial access coverage • MSG 4 carried on PDSCH with associated PDCCH

(not used in non-standalone deployments)
• Downlink grant carried on PDCCH
The effective cell range is thus determined by which of the
• HARQ ACK/NACKs carried on PUCCH Data on PDSCH supporting channels that break first. A coverage enhancement
• CSI carried on PUSCH/PUCCH feature which provides a coverage boost of the most limiting
channel will thus increase the effective cell range so that it is now
• RLC ACK/NACK carried on PUSCH determined by the next limiting control channel, as will be further
discussed on the next page [TP7.1, p. 95].

Link budgets further explained

Link budgets further explained N
EIRP
Key parameters affecting coverage include:
Tx antenna gain
• Massive MIMO and antenna gain Transmit power
• Different physical channels have different SINR requirements Total loss
• Product impact: Transmit power, receiver sensitivity, bandwidth Wall loss, etc
and duplex operation
Received Power
Rx antenna gain
EIRxP
Example mid-band link budget
170 163,6 163,3 164,6
161,5 161,5
Max supported loss [dB]
165 161,1 160,6

160 157,0
155 150,6
150 144,8
145
140
135
130
SSB PRACH MSG2 MSG3 PUCCH PDCCH DL Data UL Data CSI feedback CSI feedback
10 Mbps 1 Mbps 10 bits 100 bits
Different products and physical channels have different link budgets (assumptions matter)
Important to improve coverage limiting channels
The link budget is a useful tool to estimate coverage, measured Some aspects to consider:
as maximum isotropic loss that can be supported. Link budgets
are useful both for uplink and downlink. Both base station and Different physical channels make use of different types of
UE performance matter. Maximum supported isotropic loss beamforming and have different requirements receiver
depends on several parameters, viz. Tx power, Tx antenna sensitivity.
gain, Rx antenna gain, Rx sensitivity and possibly also some
Link budgets often consider boresight, i.e. in the direction where
other parameters, e.g. power booster factor. Different physical
the main lobe takes its maximum value. With a Massive MIMO
channels have different parameter values and hence different
you can realize the UE-specific gain over a larger area compared
coverage.
to a conventional antenna
Max isotropic loss is calculated as (simplest version):
Link budgets can be used for both uplink and downlink.
Max_iso_loss (dB) = Tx power + Tx antenna gain + Rx antenna The parameters will of course be different, but the principles are
gain – Rx sensitivity the same. Often, the uplink is limiting, since the UE output power
is limited according to standard, whereas the power from the
And is used to assess the maximum coverage for different remote radio or Massive MIMO can be typically 30-50 dB higher.
physical channels.
Since all different channels have different coverage, it is
important to identify which channel is limiting and improve
coverage of this channel first. The coverage enhancing features
typically address this issue.

Ways to improve access and app coverage

Ways to improve access and app coverage
Example: Flexible DL Tx beamforming of PDSCH PDSCH

Improving SNR using codebook feedback
• Transmit beamforming Effect: Ensure full beamforming gain and maximized
• Receive beamforming SNR in the entire cell coverage area to improve
PDSCH performance
• Power boosting
Requirements: UE CSI reporting capability, UCI
• Increase TX power coverage
Reducing interference (improving SINR) Example: UE-specific beamforming of DL control
• Interference rejection combining (IRC) signaling such as PDCCH
• Inter-cell interference planning Effect: Improved SNR and thus coverage and range of
PDCCH
• Interference avoidance
Requirements: Improved SNR and thus coverage and PDCCH
Improving detection reliability given an SNR range of PDCCH
• More robust coding format Example: Full-dimension UL Rx beamforming and
• More robust waveform full IRC of PUSCH PUSCH
• Repetition/Retransmission Effect: Rx beamform to improve signal of target user
while Rx null form to suppress interference, to
Spectrum interworking improve SINR of PUSCH
Interfering
• Transmitting of the signal on lower band Requirements: Advanced base station Massive MIMO
user
processing
There are many ways to improve access and app coverage of All of these methods are generally applicable to both
a mobile network, but on a high-level it is achieved by either conventional radios with few radio chains as well as Massive
boosting the signal strength of a signal/channel or reducing the MIMO radios. However, with transmit/receive beamforming and
interference level at the receiver. Boosting the signal strength IRC receiver processing, the benefit increases with the number
(or SNR) can either be done directly by increasing the transmit of radio chains and so these methods can be considered as
power, maintaining the same power level but boosting the PSD differentiators for Massive MIMO systems to further improve the
of certain resource elements in frequency while de-boosting coverage compared to conventional radio systems.
others, or performing beamforming on the transmit or receive
side. Or, indirectly by improving the detection reliability for a
given SNR (and thus increasing the energy per bit), such as
by performing repetition or choosing a more robust coding/
transmission format. Data channel coverage can also be
improved by reducing the interference level (in interference-
limited scenarios) by either inter-cell interference planning of
the transmission, or by utilizing advanced receiver processing
such as IRC. Finally, one can improve the coverage of a signal
by choosing to transmit it on another channel with better
propagation conditions such a lower frequency band with lower
path loss.

Flexible versus restricted beamforming

Flexible versus restricted beamforming N
Ch
Single wide
beam
No dynamic beamforming (wide beam) Restricted beamforming Flexible beamforming

• Single wide beam • A few wide beams covering entire cell service • Many narrow candidate beams covering
area or a few narrow beams covering subset entire cell service area
• Tradeoff between beamwidth and antenna of cells service area
gain • Full beamforming gain in all directions
• Good coverage where beams are pointing
• Baseline coverage (similar to non Massive
(typically boresight), poor coverage in other Candidate beam
MIMO due to lack of directive beamforming) directions Intended cell
• Reason: Broadcast information or where UE service area
• Reason: Limited CSI availability Selected beam
channel not known
Restricted and flexible beamforming can offer different coverage versus overhead tradeoffs
Even though a Massive MIMO radio has the possibility to use For instance, in order to apply flexible beamforming of the
classical beamforming to create narrow beams maximizing the PDSCH for a 32Tx Massive MIMO, CSI needs to be available
beamforming gain in a certain direction, it also has the possibility either from uplink sounding on or 32 port CSI-RS feedback.
to create beams with larger beamwidth (and thus lower antenna However, some UEs may only support 8 CSI-RS ports and so a
gain). The benefit with using wider beams is that less beams are restricted beamforming with a fixed narrow beam in the vertical
required to cover the entire intended cell service area, which is domain but many narrow beams in the horizontal domain can be
some cases may be desirable, on the other hand the downside is applied for that UE. In this case, the restricted beamforming still
that the experienced beamforming gain is lower. results in narrow beams with maximum beamforming gain but
does not cover the entire cell service area in the vertical domain
In general, it is most beneficial for coverage to apply flexible and so a UE which is not in the vertical direction of the fixed
beamforming, with many narrow candidate beams covering vertical beam will see limited beamforming gain.
the entire cell service area, whenever possible in both downlink
and uplink in order to maximize the beamforming gain and thus Another way to apply restricted beamforming is to instead create
improving the coverage. However, there are some exceptions a set of wide beams which covers the entire intended cell service
where restricted beamforming or even a single wide beam may area, but with lower maximum beamforming gain per beam.
be more suitable: In the extreme case of this variant of restricted beamforming,
we end up with a single wide beam covering the entire cell
1. Additional coverage improvement is not necessary service area. This way of beamforming may be suitable
2. Lack of CSI availability for broadcasting of cell-defining signals such as SSB and
broadcasting of system information.
Note: The beams are for simplicity depicted as classical beams.

In reality, however, they should viewed as generalized beams
[TP3, p. 27].

Example: Access coverage

improvement
Example: Accesspath
coverage
improvement path
Graphical representation of 2D
0. Baseline System
physical channel coverage
The thick black line represents the DL data CSI feedback
overall mid-band coverage defined PDCCH UL data
by the worst physical channel Overall mid-band coverage
1. Moving UL data to LTE with 2. Moving CSI to NR FDD with 3. Improving PDCCH with beamforming
UL/DL decoupling in EN-DC inter-band DL CA
We here provide one illustrative practical example of how overall To further improve the coverage, inter-band downlink carrier
mid-band TDD coverage can be improved by a set of coverage aggregation can be activated to move the uplink control
enhancement features. channel (carrying the CSI report) also to low-band FDD, thereby
increasing the range where CSI reports can be correctly decoded
In the baseline system, no spectrum interworking is used, and increasing the effective cell range further, which is now
and all channels are transmitted on the mid-band TDD carrier. limited by PDCCH.
The effective cell range is here limited by the uplink data channel
(PUSCH). User-specific beamforming of PDCCH can be introduced to
beamforming PDCCH in a similar fashion as PDSCH.
By using spectrum interworking with low-band FDD in EN-DC,
the uplink leg can be moved to LTE on the lower FDD frequency
band with better coverage properties, thereby removing the
first coverage limitation. The effective cell coverage is now
determined by the CSI feedback coverage of the UCI.


7.2 Capacity and

performance

Overview of Massive MIMO features

for downlink
Overview data MIMO
of Massive channel
features for N
downlink data channel W
SU-MIMO MU-MIMO
CSI-RS
Codebook- CSI-RS CSI-RS
based
k
ac ck
db ba CSI feedback
I fe e fe e
d
CS I
CS
SRS
Reciprocity-
based
SRS
S
SR
S
SR
SR
On a high level, the basic bread-and-butter features for downlink

performance and capacity in a Massive MIMO system can be
classified according to 1) if codebook-based or reciprocity-
based channel information is used and 2) whether single- or
multi-user MIMO is applied. This results in four distinct high level
downlink Massive MIMO features. Typically, all of these four
features would be activated in a cell, but which transmission
scheme is used at a certain point in time would depend e.g.
traffic conditions and UE characteristics. Of course, there are
various different possible options for how reciprocity-based
and codebook-based SU/MU-MIMO can be implemented, both
from what is available in the 3GPP standard as well as from
proprietary algorithm perspective, however on a high-level the
various options still share the same characteristics with respect
to use case and performance.

Codebook-based beamforming
Codebook-based beamforming Change
CSI-RS
CSI feedback
With codebook-based transmission, the downlink channel However, since the feedback format is limited (the precoder
state is acquired by first sounding the channel in the downlink codebook can be seen as a course spatial quantization only
by the base station transmitting a Channel State Information giving information about the strongest propagation path of the
Reference Signal (CSI-RS) which is measured on by the UE. The downlink channel), only simple classical beamforming, [TP2,
UE thus has full information about the downlink channel state p. 15], can be used. Such beamforming works quite well for
but needs to convey this information to the base station in order low-rank transmission in line-of-sight channels, but generalized
for it to determine in which direction to beamform. Theoretically beamforming requiring more detailed channel information is
this could be achieved by simply quantizing all the channel typically needed for good performance with high-rank (rank >2)
coefficient to a number bits and transmitting these in the uplink, transmission or when the channel contains a lot of multi-path.
so called explicit feedback, however this would result in massive As the codebook-based feedback only gives the base station
overhead. Instead, a so called implicit feedback mechanism is information about the dominant downlink channel direction,
used, where a number of candidate beamformers are defined in it is more suitable for SU-MIMO transmission. MU-MIMO with
the 3GPP standard, a so called precoder codebook, where the UE codebook-based feedback is still possible and can give a benefit
recommends the most suitable beamformer for the base station in some cases (especially with well-separated users or with
to use, i.e. the one that best matches the measure downlink sparse channels without a lot of multi-path), but will in the
channel on the CSI-RS. The number of candidate beamformers general case not give substantial gains since the interference
defined in the 3GPP codebooks are relatively few, up to ~256, so between users is difficult to suppress.
the selection can easily be encoded into a few bits and sent with
channel coding in the uplink control information, making the
coverage of such feedback quite robust.

Reciprocity-based beamforming
Reciprocity-based beamforming N
SRS
With reciprocity-based beamforming, the downlink channel However, in order to use the fast-fading properties of the
state is instead acquired by sounding the uplink channel by channel for precise nullforming and beamforming, the channel
the UE transmitting a Sounding Reference Signal (SRS). By estimates needs to be of sufficiently good quality. This means
utilizing that the uplink propagation channel is reciprocal since that coverage range where such transmission can be performed
the uplink and downlink is on the same frequency in a TDD is limited. Trying to conduct precise nullforming/beamforming
system, the downlink channel state can be estimated from based on too noisy channel estimates generally results in
the uplink (while the propagation channel is reciprocal, the worse performance than what can be achieved with classical
channel effects due to RX/TX processing is generally not, so beamforming using codebook-based feedback. Another
calibration between the uplink and downlink chains is required limitation with reciprocity-based beamforming is that the entire
for reciprocity-based beamforming to work). The benefit with bandwidth to be scheduled in the downlink needs to be sounded
reciprocity-based beamforming is that the complete channel in the uplink, for each Rx antenna of the UE. This means that
information is available at the transmitter, including it’s fast the UE needs to have an uplink allocated on the TDD cell (and
fading properties, which gives the base station very granular so reciprocity-based beamforming may not be available in a
channel information in spatial and frequency domains enabling downlink carrier aggregation scenario). And since mid-band NR
the use of more advanced generalized beamforming, [TP3, p. UEs typically only are equipped with 1 or 2 Tx chains but have
27]. The generalized beamforming can be used to increase both 4 Rx antennas, the proper SRS antenna switching functionality
SU-MIMO performance by creating high-rank beams which needs to be supported by the UE so that all Rx antennas can
correspond better to the actual propagation channel (no longer be sounded. Each user also needs to get allocated a dedicated
limited by the fixed feedback format of the 3GPP codebooks), SRS resource. Since the SRS resources are limited from an air
and particularly to boost MU-MIMO performance since the more interface perspective, not all users may have the possibility to
detailed channel information can be used for nullforming to have an SRS resource assigned if many users are connected to
reduce interference between UEs. a cell and hence SRS resources should be prioritized to the users
that would benefit from reciprocity-based beamforming the
most at a certain point in time.

Overview of features and benefit

for downlink data channel
Overview of features and benefit for No chan
downlink data channel Wrong fo
SU-MIMO MU-MIMO
Coverage Coverage
Codebook- Capacity Capacity
based Peak rate Peak rate
Dynamic Codebook/reciprocity and SU/MU switching

Coverage
Capacity
Peak rate
Reciprocity- Coverage Coverage

based Capacity Capacity
Peak rate Peak rate
If the four Massive MIMO features are compared with respect + Good low-rank (1-2) SU-MIMO performance
to our network key performance indicators of interest: coverage,
capacity and single-user peak rate we can see that they – Limited high-rank (3-4) SU-MIMO performance
have different strengths and weaknesses. Codebook-based
– Poor MU-MIMO performance when channels are overlapped
beamforming has an advantage in coverage over reciprocity-
based beamforming as was discussed on the previous slide. Reciprocity-based beamforming
Similarly, SU-MIMO has a coverage advantage over MU-MIMO
as the TX power needs to be split between multiple users in + Better high-rank SU-MIMO and MU-MIMO due to
the latter case. When it comes to capacity, reciprocity-based improved inter-layer interference suppression capability
MU-MIMO can provide the most (but since it is coverage due to spatially rich channel information
is limited, the maximum benefit can only be seen in dense
deployments with small cell sizes so that most of the users in the – SRS resource limitations
cell is within coverage).
• Each user needs dedicated SRS resource(s)
To fully utilize to potential of a Massive MIMO system, it is thus
• Need to sound the entire bandwidth to be scheduled
necessary to dynamically switch between all of the four above
schemes so that coverage capacity and peak rate jointly can be • Allocated uplink needed, either as PCell or using SRS carrier
maximized, which is also how the typical Massive MIMO systems switching
are designed.
• Each Rx antenna needs to be sounded (e.g. 1T4R SRS Tx
switching capability needed)
Codebook-based beamforming – SRS coverage limitations
+ Superior coverage of the CSI acquisition, since only • Good uplink channel estimation quality needed → limited
wideband channel information needed (e.g. ~10 bits coverage
PMI feedback)
+ Coverage extension possible with inter-band carrier

aggregation by sending CSI report on low-band FDD
+ MU-MIMO for well separated users

In which scenario is each scheme

most useful?
In which scenario is each scheme most useful? N
Reciprocity-based
MU-MIMO
Codebook-based
Traffic demand
MU-MIMO
Reciprocity-based Codebook-based
SU-MIMO SU-MIMO
Good coverage Poor coverage
We can visually map out where the different schemes would

typically be applied on a two-dimensional grid with the traffic
demand, e.g PRB utilization, of the cell on the y-axis and
coverage on the x-axis.

Codebook-based beamforming options

Codebook-based beamforming options No chan
CSI-RS options CSI-RS illustration Resulting beam granularity
Cell-specific full-dimension CSI-RS 32 ports
• E.g. Single CSI-RS resource with 32 ports
4 CSI-RS, 8 ports each

Cell-specific sub-sector CSI-RS
• E.g. 4 beamformed CSI-RS resources

transmitted with 8 ports each
CSI-RS A CSI-RS B CSI-RS C CSI-RS D
1 CSI-RS, 4 ports
UE-specific beamformed CSI-RS
UL signal
• E.g. single CSI-RS resource with 4 ports N/A
The 5G standard is flexible and generally offers multiple Full-dimension and sub-sectorized CSI-RS
implementation choices. Specifically for codebook-based Cell-specific CSI-RS transmission can further be broken down
beamforming, there is a variety of different ways it may be into full-dimension or sub-sector CSI-RS transmission. With full-
realized. Three common implementation choices are illustrated dimension CSI-RS, one CSI-RS resource with many ports, e.g.
here. The CSI-RS transmission can be either cell-specific or 32, are used. For a 32 Tx Massive MIMO radio, one CSI-RS port
UE-specific. can thus be transmitted from each Tx chain, enabling sounding
of the full channel in downlink. The UE is thus aware of the full
Cell-specific and UE-specific CSI-RS dimensional propagation channel and can select a precoder from
With cell-specific CSI-RS, a single CSI-RS resource or set the 32-port codebook ,which corresponds to 256 hypothetical
of resources are broadcasted to all the UE in the cell, i.e. the beam directions covering the entire channel space. With a sub-
UE’s share a common set of CSI-RS resources, whereas with sector based CSI-RS transmission strategy on the other hand,
UE-specific CSI-RS each UE gets its own CSI-RS resource, multiple beamformed CSI-RS resources, each with a smaller
beamformed towards that UE. UE-specific beamformed CSI-RS number of CSI-RS ports, such as 8 ports, is used. For instance,
is typically used when the base station knows the beam direction 4 CSI-RS resources, each with a different beamforming can be
towards the UE already based on some other information, used to cover a different vertical or horizontal sub-sector of the
e.g. uplink sounding, and the base station just wants CQI and cell. The UE may be dynamically or semi-statically allocated to a
RI information to aid the link adaptation. With cell-specific CSI-RS resource and the UE would report a PMI from the 8-port
CSI-RS on the other hand, the intention is for the base station codebook, corresponding to a selection from 64 hypothetical
to also determine the beam direction. The RS overhead for beams within the beamforming pattern for the sub-sector. Each
cell-specific CSI-RS is static and does not scale with the number sub-sector thus corresponds to a subset of the full channel space.
of UEs connected to the cell, since all UE share the same CSI-RS Depending on how the base station selects the beamforming
resource, however for UE-specific CSI-RS the beam direction weights of the sectors, the sub-sector may be overlapping or not.
for each UE needs to be tracked and each UE requires its own
CSI-RS resource.

Example: Beamforming/Carrier Aggregation

(CA) mode
Example: switching Aggregation
Beamforming/Carrier
(CA) mode switching
FDD UL FDD DL TDD UL TDD DL Not activated
TDD only SRS SU/MU-MIMO TDD only SRS SU/MU-MIMO FDD-TDD DL CA with FDD SpCell
1 CC FDD + 1 DL CC TDD CA
Mid-band TDD Mid-band TDD Mid -bands TDD
Reciprocity SpCell Codebook SpCell Codebook SCell
Low/Mid-band FDD Low/Mid-band FDD Low/Mid-band FDD

SpCell
Reciprocity-based Codebook-based with CSI feedback Codebook-based with CSI feedback

on TDD SpCell on FDD SpCell
Good RF Mid RF Poor RF
Coverage
As what kind of beamforming can be used depends on what When the UE moves out towards the cell edge into the mid-RF
kind of CSI is available to the base station, it is important to condition region, the uplink SNR may be so poor that the quality
consider the interaction with Massive MIMO feature together of the SRS channel reception is not sufficient to provide good
with other RAN features such as Carrier Aggregation. Different reciprocity-based beamforming performance. Instead, the base
combinations of Massive MIMO and carrier aggregation features station falls back to codebook-based beamforming for this UE
may be activated in different parts of the cell depending on for and may also deallocate this UE’s SRS resource and give it over
example coverage and other factors. to another UE in better RF condition.
Consider the above example with one TDD carrier and one As the UE moves further out towards the cell edge into the poor
FDD carrier. Close to the cell center when the UE is in good RF RF condition region, the TDD uplink quality may be too poor
conditions, it is beneficial to set the TDD cell as a SpCell so that to even carry the CSI report in the uplink control information.
the uplink is allocated to the TDD carrier (the FDD cell may then At this point, the base station would activate downlink carrier
be inactivated as in this example or set as an SCell). Since the aggregation and configure the FDD carrier as the SpCell (or
TDD uplink is in good coverage, SRS can be received with good simply make a SpCell change if downlink carrier aggregation is
quality and the UE can use reciprocity-based SU/MU-MIMO. already activated with a TDD SpCell), thereby moving the uplink
to the lower band with better propagation conditions, leaving the
TDD carrier as a downlink only cell. The UE can still measure the
CSI-RS on the TDD Scell and determine the preferred codebook
beam, but it sends the CSI report with the indication on the
uplink of the FDD SpCell, thereby increasing the coverage range.
Examples of the benefits of frequency inter-working are provided

in [TP9, p. 159-160], where it is shown that the coverage using
carrier aggregation with low band FDD can improve coverage by
16 dB.

MIMO features for uplink data channel
MIMO features for uplink data channel No chan
Uplink SIMO Uplink SU-MIMO (2 Tx) Uplink MU-MIMO
1 Tx UE 2 Tx UE
Similar to the Massive MIMO features for downlink data channel, 2. Commercial Ues are typically only equipped with either one or
SU- and MU-MIMO can be applied on the uplink data channel two (non-coherent) Tx chains and corresponding antennas,
as well. However, there are some fundamental differences which means that the precoding operation on the UE is quite
compared to the DL. simple: transmit one layer from one antenna or two layers,
each from a separate antenna / Tx chain. This means that
1. Since the Massive MIMO processing is performed at the the base station does not need to provide any CSI to the UE
receiver side after transmission (i.e. not before transmission either, but can simply indicate transmission rank and possibly
as for downlink MIMO), the base station does not need to antenna selection (although the 3GPP standard defines more
acquire CSI but can directly base the Rx beamforming/ advanced uplink transmission modes, these are not expected
nullforming processing on the DMRS transmitted with the to be implemented on a larger scale).
PUSCH.
3. Resulting uplink performance depends mainly on the receiver
processing at the base station

Digital sectorization (FDD)
Digital sectorization (FDD) N
Digital sectorization is a method to increase the capacity on site The increase in capacity is somewhat dependent on the network
when the penetration of LTE terminals with TM9 or better is not topology and diminishing for every split but typical capacity
high enough to allow for multi-user MIMO. improvements are in the order of 50% for a split by two and 70%
for a split by three.
By using an FDD Massive MIMO radio, a conventional 120
degree sector is split into two or more narrow sectors that can
simultaneously address users. Effectively, this increases the
number of PRBs available in the original coverage area. As this
feature works with any kind of terminal, it provides a very robust
way of increasing the capacity in an LTE network.

7.3 Advanced MIMO

for extreme performance

High-order MU-MIMO
High-order MU-MIMO N
High-order MU-MIMO Extreme capacity demand

(>8 layers MU-MIMO) SRS
SRS
High-capacity demand
(≤ 8 layers MU-MIMO)
Low-Mid capacity demand
(SU-MIMO)
Traffic per cell
S
SR
SRS
5% SRS
S
SRS
SR
20%
50% SRS
Cells
100%
While MU-MIMO is a key feature to provide capacity in a Massive However, the high-capacity Massive MIMO radios are
MIMO system, the number of MU-MIMO layers required to fulfill dimensioned to support 16 downlink layers and 8 uplink layers,
the capacity demand is generally quite few in practice. In fact, since this could be required in certain cell sites with extreme
in a large portion of the network the capacity demand can be capacity demands, such as traffic hotspots or stadium/arena
fulfilled with only SU-MIMO. Even for cells with high capacity deployments with a lot of people in a small area, or non-MBB use
demand, it is generally sufficient with 8 downlink layers and cases such as FWA. Higher-order MU-MIMO can also be useful
4 uplink layers to reap most of the MU-MIMO gains with MBB for scenarios with different traffic characteristics than bursty
traffic. packet-based MBB traffic. MU-MIMO is further elaborated on in
[TP3, p. 33, 40-41] and [TP9, p. 157].

Coordinated multi-point (CoMP)

Coordinated multi-point (CoMP) No chan
Wrong fo
Coordinated beamforming (CBF) Coordinated scheduling (CS) Joint transmission/reception (JT/JR)

Avoids inter-cell interference Avoids inter-cell interference Increases signal level
Coordinate nulls towards victim UEs Coordinate resource allocation May increase or decrease interference
Coordinate link adaption JT: Multiple points/cells transmit to a UE
JR: Multiple points receive from a UE
Nul
l
Coordination Coordination
Coordinated multi-point transmission/reception, or CoMP, Coordinated beamforming: With coordinated beamforming,

utilizes cooperation and coordination between multiple cells (or the intention is for reduce the generated inter interference by
transmission/reception points, TRPs, of the same cell) to improve adapting the beamforming pattern so that neighboring cell UEs
network performance, primarily by either aiming to reduce are not struck by interference. Typically, this can be achieved by
inter-cell interference or utilizing the transmission/reception generating nulls towards the victim UEs in the neighboring cells,
resources of multiple TRPs to boost the signal energy of a user. similar to how nulls are generated towards the other layers/UEs
Many different flavors of CoMP exists, with different pros and in a MU-MIMO scheduling. Hence, coordinated beamforming
cons, different requirements on backhaul/coordination and can be seen as an extension of MU-MIMO nullforming also
applicability to different scenarios. considering neighboring cell UEs.
Coordinated scheduling: One set of CoMP features utilizes Joint transmission/reception: For coordinated scheduling and
coordination on the scheduler level to for instance adapt beamforming, the UE is only receiving transmissions from a
the resource allocation between TRPs to reduce inter-cell single TRP at a given time, in joint transmission (JT), the UE can
interference. For instance, the scheduler can make a dynamic receive transmissions from multiple TRPs simultaneously. JT
decision to not schedule any UEs from one or more TRPs, thereby can further be broken down into two flavors, non-coherent JT
reducing interference to the UEs served by the remaining TRPs. (NC-JT) and coherent JT (C-JT). For C-JT, the same precoding
Another flavor of coordinated scheudling is coordinated link layer(s) are transmitted from the multiple cooperating TRPs with
adaptation, CoLA, where each TRP makes scheduling decisions the intention for the transmissions to coherently add up at the
independently, but exchange information between one receiving UE. In order to accomplish this, tight synchronization
another so that the link adaptation can take information about is required between the TRPs and a high CSI accuracy is needed
interference from the neighboring TRPs into account. in order to design the precoding weights, which makes C-JT
difficult to practically implement and requires large CSI feedback
overhead. In NC-JT, on the other hand, the requirement for
synchronization and CSI accuracy is comparably lower, since
each TRP is transmitting different layers.
Analogously, with joint reception (JR) the received signal (either

before or after demodulation and mapping to soft bits) from
multiple TRPs can be accumulated, hence boosting the signal
level and increasing decoding performance.

Interference Sensing
Interference Sensing
Superior Massive MIMO architecture Ericsson

New
Uplink
Booster Interference
Sensing
Access to the full-dimensional channel for uplink processing technology
Improved ultra-precise beamforming
Interference Inter-cell
Sensing nullforming
….. Beamforming
….. Nullforming
As traffic demand increases in a mobile network, its performance Ericsson Interference Sensing analyzes the network
and capacity becomes limited by inter-cell interference. Thus, surroundings to obtain interference characteristics and generate
by reducing inter-cell interference, performance degradation at an interference profile. These characteristics can then be
higher traffic loads can be mitigated and network capacity can exploited in combination with state-of-the-art beamforming to
be increased. avoid emitting interference towards neighboring cells.
Massive MIMO radios can utilize spatial interference avoidance Based on the sensed interference profile together with the
techniques, such as coordinated beamforming, to distribute the channel estimates for the serving cell users, a set of beamforming
generated signal energy to where it is least harmful to the users weights are generated – jointly optimized towards avoiding
in neighboring cells. However, especially in denser networks, one to generate interference to the neighbor cell users while
cell can generate significant interference to a large number of simultaneously maximizing the signal energy of the serving
neighboring cells and furthermore different users can be active cell users – using advanced state-of-the-art signal processing
at different times in those cells, making interference avoidance techniques. These calculations are extremely processing heavy,
a difficult problem. Traditional coordinated beamforming requiring around 10-40 billion calculations per second, but can
techniques relies on tight coordination and information easily be realized on the powerful Ericsson Silicon compute
exchange between schedulers over the backhaul, which limits platform.
the applicability to a small number of nodes in practice. To
circumvent such limitations, the Interference Sensing feature By deploying Massive MIMO and applying Ericsson’s state-
realizes coordinated beamforming by sensing over-the-air which of-the-art Interference Sensing beamforming technique, the
reduces the need for tight scheduler coordination. inter-cell interference towards a large number of neighboring
cell users can be reduced, which improves user performance
The Ericsson unique Massive MIMO architecture does not only at higher traffic loads and substantially increases the overall
enable access to the full-dimensional channel for the uplink network capacity, see [TP9, p. 158].
receiver processing of our UL Booster, but also enables improved
downlink beamforming with the new Interference Sensing
technology.

Intelligent Cell Shaping

Intelligent cell-shaping No chan
Possibility to tailor the cell shape

• Shape the beam(s) of cell defining signals such as SSB in NR
• Slow update rate – static or semi-static
Benefits
• Lower inter-cell interference – avoid cell-edge in hotspot
• Raise the signal level – point where the UEs are
• Load balancing – move UEs from one cell to another
Artificial intelligence (AI) and machine learning

(ML) application
The cell shape depends on the beam shape of the cell-defining Cell shaping can be used to solve various problems. For example,
signals, such as the SSB in NR. This provides the possibility for the cell-border with its interference impairments can be moved
the network to tailor the cell shape for some specific purpose to avoid being in the middle of a location where typically there
by changing the beam shape and slowly update the cell shape are many simultaneously connected UEs, known as a hotspot.
instead of striving for fixed cell areas with potentially hexagonal An example of such a hotspot is a train station. Using cell shaping
structure. The beam shape of a cell defining signal is controlled to avoid a cell border within a hotspot is illustrated in the figure.
by the use of a beamforming vector (1) that maps the signal Cell shaping also offers the possibility for load balancing among
onto the antennas, just as in the dynamic beamforming case. cells by moving UEs from one cell to another when altering the
By measuring various parameters in the network, suitable uptake areas of cells.
beam shapes can be determined. The technique of letting the
system adapt and update the beam shape of cell-defining Cell shaping can also reduce interference in unwanted directions.
signals is called cell shaping. Cell shaping can be used together Such a reduction is especially important in transmit directions
with dynamic and UE specific beamforming to further improve that negatively impact particularly many UEs in neighboring
performance or it can be used stand alone. cells.
Intelligent Cell Shaping makes use of advanced AI/ML

algorithms to leverage from the very large data volumes
collected over long periods of time.


7.4 Ease-of-deployment,
cost and energy
efficiency

Directional interference mitigation

Directional interference mitigation
Nullforming
Some mid-band TDD bands in certain regions have spectrum Spatial nulls towards the sensitive directions can be inserted in
coexistence requirements with regulatory limitations on both the broadcast beam, CSI-RS beam as well as PDSCH beam.
generated interference towards spectrum incumbents, such
as satellite earth station receivers, which may be operating The interference generated towards particularly sensitive
either in-band or on adjacent band. The Directional interference directions can be reduced, in order to comply with regulatory
mitigation feature enables limiting the radiated power of requirements, without reducing the signal level in other
a Massive MIMO in certain sensitive directions, such as the directions.
direction of an earth station receiver, by utilizing spatial
Broadcast beam pattern can be adopted to change the cell
nullforming. This enables the radiated power to be reduced only
uptake area to not attach users in the sensitive directions.
in those directions required, but full power to be utilized in other
non-sensitive directions, to maximize the coverage and capacity
of the cell.

Electromagnetic Field (EMF) power lock

protection
Electromagnetic Field (EMF) power lock protection
On Massive MIMO solutions, the instantaneous Exclusion zone for 10 W/m2

ICNIRP public EMF limit
EIRP and the resulting peak RF EMF
General public
levels can be higher than those
Exclusion zone for 50 W/m2
of traditional BS antennas Massive MIMO
Workers ICNIRP public EMF limit
Legacy
technology
Instantaneous
EIRP
Average total power Average power Control Dynamic threshold Data
Radio
controller signal scheduler actuator
Total power
computation
Total average
UE power estimation
The use of Massive MIMO solutions means the If time-averaging is not considered, the size of the RF EMF
instantaneous effective isotropic radiated power (EIRP) and exclusion zone increases when the BF gain increases. Such an
the resulting peak (non time-averaged) radio frequency (RF) increased exclusion zone could make deployment challenging in
electromagnetic field (EMF) levels can be higher than those e.g. dense urban environments. IEC TR 62669 - is recommending
for traditional base-station antennas. The RF EMF exposure is monitoring and control of the time-averaged power, to a
regulated in most markets, the RF EMF exposure limits, defined pre-determined level, to obtain a certain exclusion zone, with a
by International Commission on Non-Ionizing Radiation minimum impact on the capacity and coverage of the served cell.
Protection (ICNIRP), are typically defined as time-averaged
values, e.g. over 6 minutes. EMF Power Lock controls the average transmit power per sector
carrier allows the configuration of a power back-off threshold
ranging btw 20.6% - 100% of the max available total power by
controlling the scheduled number of PRBs for PDSCH.

NR has great energy saving possibilities

NR has great energy saving possibilities No
507 W 378 W
(average power) (average power)
LTE NR
The NR standard is considerably better compared to previous In the LTE standard, due to the mandatory transmission of
standards in terms of low-energy operation. The crucial cell specific reference signals (CRS) from the network in every
difference is the vastly improved support for energy savings subframe, these time gaps are always very short, at the most
during low-to-medium traffic. NR also has much higher capacity, 0.2 ms. In NR, the time between mandatory transmissions from
resulting in a lower load in terms of the fraction of the capacity the network (i.e. the SSB) can be 20 ms in stand-alone mode and
that a certain traffic volume utilizes. 160 ms in non-stand-alone mode. That’s 100 to 800 times longer
than the corresponding value of 0.2 ms for LTE. In addition,
When examining the traffic patterns in the networks, several NR requires far less always-on signaling transmissions in the
short gaps in the data transmissions can be observed, even frequency domain. This is often referred to as the ultra-lean
during highly loaded times. These short gaps are utilized to design of the NR physical layer and was a key design criteria.
reduce power consumption in the network by quickly putting It is the ultra-lean design that allows for both deeper and
components into sleep mode and then activating them again longer periods of sleep when there is little or no ongoing data
before the next transmission occurs. The longer these time gaps transmission.
are, the more components can be put to sleep, and the lower the
energy consumption becomes. The graphs show power measurement in two different modes:
To the left (in orange) show the power-versus-time signature of
a 5x20 MHz LTE system and to the right (in green) a 100 MHz
NR system is depicted. Both examples use 200 W output
radio frequency (RF) power. The reduction in average power
consumption (a saving of 129 W) in NR is due to the ultra-lean
design of NR and the much more effective use of micro sleep TX.

Energy efficiency Massive MIMO features

Energy efficiency Massive MIMO features No chan
Zooma
of the le
Booster Carrier sleep: Disable booster carriers User data booster Carrier Traffic booster Carrier Basic Carrier
Frequency
Mic
ro/
Mil
li Sl
e ep
Tx Massive MIMO sleep:
Disable parts of the antenna
Massive MIMO solution provides high capacity and performance Different features, handling the different perspectives,
which is useful for handling high traffic load scenarios such as collaborates to provide the required service level of capacity
repeating busy hours and also occasional peak loads. and performance. Three types of features are in the solution:
In most traffic load scenarios, even in the most loaded areas, • Micro/milli sleep Tx, autonomous discontinuous transmission
there are still periods with low traffic load where the available features, act to set different components in radio PA in sleep
capacity is not required. During these periods power can be based on empty slots in transmitted down link. No impact on
saved by deactivating not required capacity. By switching off any key performance indicator.
radio unit components power consumption is reduced.
• Frequency-domain based Booster Carrier Sleep that
Different Massive MIMO radios have different capability in terms deactivates not required carriers. All available carriers are
of what HW components that can be switched off. The RAN tagged as Basic or Traffic Booster or User Data Booster carrier.
SW part of the solution is agnostic to different radio units. This Basic carrier is the carrier that will transmit necessary cell info
enables a feature such as Massive MIMO Sleep to work with any such as system information, thus providing the cell coverage.
Massive MIMO radio. It will always be on. The two Booster carriers will be set to sleep
based on need.
Apart from Massive MIMO specific energy saving solutions there
are also RAN level solutions that for each coverage area adapts • Massive MIMO Sleep is a Massive MIMO radio specific feature
provided capacity to the actual needed by sensing and predicting that will deactivate (set to sleep) not required segments.
traffic. Coverage is maintained.
All features can be activated and they will inter-work to find

the best power savings. In addition to have even more savings
Energy Performance Optimizer (EPO), a central AI based tool,
can orchestrate the different energy savings features to find
even more savings. By using advanced traffic sensing and
prediction functionalities EPO can find the periods for saving
and also decide the most efficient sleep level for the equipment.
The deeper sleep the longer activation time. Deeper sleep means
more parts of the radio unit will be deactivated.

Massive MIMO sleep

Massive MIMO sleep N
Le
Transmission without Massive MIMO sleep mode Transmission with Massive MIMO sleep mode
4x4 antenna layout with 32 antenna ports

4x4 antenna layout with 32 active antenna ports of which 16 antenna ports are active
(active ports in blue) (active ports in blue and muted ports in grey)
NR with Massive MIMO provides very high peak capacity, and In the example above a 32 TX Massive MIMO antenna array
potentially an increase in the total network energy consumption consisting of 16 dual polarized antenna elements arranged in 4
as a result of new deployment of additional equipment. Since rows and 4 columns are deployed. When Massive MIMO sleep
the traffic load varies significantly over time and space the high is activated, only a subset of these antenna elements remain
peak capacity provided by NR with Massive MIMO is not needed active. In the example above, two columns of dual polarized
everywhere and all the time. This temporal and spatial over- antenna elements are deactivated, effectively reducing the
capacity creates opportunities to reduce the network energy number of active antenna elements to 16 TX enabling operation
consumption. Massive MIMO sleep is a technique that can be with reduced energy consumption. A machine learning (ML)
used to scale down the active antenna array size when traffic algorithm is used to trigger the switch between “normal Massive
demand is low, hence reducing energy consumption. MIMO operation” and “Massive MIMO sleep operation”.

Deep sleep for Massive MIMO

Deep sleep for Massive MIMO No chan
Active digital radio parts Digital radio parts in deep sleep
An effective way to reduce power consumption of a Massive

MIMO Radio in a low traffic scenario is to use Ericsson ASIC
advanced power saving techniques. The figure shows a heat
map of the Massive MIMO digital radio. Red and blue colors
denote high and low temperatures, respectively. On the left side,
the Massive MIMO radio ASICs are still activated and in use.
On the right side the ASICs now have entered a low power stage.
This Deep Sleep for Massive MIMO feature can be used in all low
traffic scenarios, e.g. during nighttime, off-peak hours in business
areas, subway stations, stadiums etc.


mmWave offers much bandwidth

but coverage is challenging
mmWave offers much bandwidth Bo
sho
but coverage is challenging
Bandwidth
Coverage challenges at higher
frequencies
wwWave Beamforming Outdoor-to-indoor Diffraction

penetration Path gain
Mid-band Beamforming
Low-band
Body loss
Foliage loss
Coverage
As discussed in chapter [Ch. 2, p. 35], mmWave is an attractive As also discussed in [TP1, p. 9], it should be emphasized that the
solution to meet the growing needs for network capacity and reason for worse path loss at higher frequency is because the
to support new services with higher performance requirements ability of an antenna to capture energy reduces with increasing
due to large amount of available bandwidth. The drawback of frequency and not due to any propagation effect. To counteract
mmWave is that coverage is substantially lower than for lower this phenomenon, a larger antenna with more antenna elements
frequency bands due to shorter wavelengths and frequency needs to be used. This will be further discussed on page [TP8, p.
dependent propagation properties [TP1, p. 9]. 128-129].
The figure to the left illustrates that mmWave has much larger In this chapter, the solutions used to improve coverage, capacity
bandwidth but lower coverage compared to other frequency and user throughput for mmWave are discussed. These solutions
bands. Massive MIMO with beamforming and interworking will cover physical hardware properties, hardware architecture,
between frequency bands are key components to improve implementation aspects and software features. Several software
coverage, and hence the usefulness of mmWave. The figure on features are similar for mid-band and mmWave. For descriptions
the right shows some coverage challenges at higher frequencies, of these features, see chapter [TP7].
e.g. lower path gain, higher penetration losses, larger diffraction
losses and increased foliage losses. Generally, mmWave The purpose of this chapter is to show the challenges that
propagation is affected more than mid-band by penetration mmWave communication presents, how these are handled and
and diffraction, but not by reflection and scattering [TP1, p. 9]. what the remaining discrepancies compared to mid-band are.
Therefore, mmWave performs better in environments dominated
by line-of-sight propagation or in environments where reflections
dominate.
mmWave numerology offers low RAN latency

mmWave numerology offers low RAN latency inteface
RAN Latency examples

Initial Tx 1st reTx 2nd reTx 3rd reTx
Numerologies:
High-band TDD, 4:1 TDD pattern, regular TTI, DL 120 kHz sub carrier
spacing (SCS)
UL
Mid-band TDD, 4:1 TDD pattern, regular TTI, DL 30 kHz sub carrier
spacing (SCS)
UL
Low-band FDD, regular TTI, DL 15 kHz sub carrier

UL spacing (SCS)
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Accumulated radio interface delay [ms]
The numerology in NR determines some fundamental parameters With vast amounts of spectrum and high numerology, mmWave
associated with the frame structure and the resource grid. NR can provide extreme data rates and low latency. Therefore,
supports a range of different such numerologies by which it is mmWave can be used as an enabler for new services, especially
possible to optimize the system for different requirements on those with high requirements on bitrate, latency and reliability.
latency or throughput.
The figure illustrates radio access network (RAN) latency
The numerology is determined by the subcarrier spacing (SCS) examples for different frequency bands using different
that ranges from 15 kHz up to 480 kHz. There is a duality numerologies. A 4:1 TDD pattern means that there are four times
between time and frequency properties such that a larger more slots allocated for downlink than uplink.
subcarrier spacing means a shorter time duration, e.g. shorter (in
time) slots. Hence, a larger subcarrier spacing with the associated There are many more aspects than just numerology that impact
shorter time slot, supports shorter layer 1 round trip delays to the end-to-end RAN latency, for example:
reduce service latency. There are several reasons why NR has
• A TDD configuration with more frequent UL/DL switching
multiple numerologies, but one important aspect relates to
such as DDSUU would reduce latency to below one millisecond
phase noise handling, see next page, which is higher for higher
(ms). Here D, U and S refer to downlink, uplink and special slots,
frequencies which imply that a larger SCS is beneficial. Another is
respectively, see [TP4].
to maintain the receiver complexity due to the FFT size, when the
carrier bandwidth gets wider the increased SCS also ensures that • Shorter TTIs would reduce latency somewhat further, but not
the total number of subcarriers of a carrier is kept reasonably low significantly. It will, however, improve capacity for small packet
(e.g,. below 4000). data transmissions.
• Time-domain beamforming may also add latency.
Radio challenges at higher frequencies

Radio challenges at higher frequencies
Power
[dBm]
Y
Y-20 dB
Frequency
x 10x
Technology challenges such as power amplifier performance arise
Massive MIMO can be used to mitigate the challenging 4. The phase noise due to frequency generation also degrades
propagation conditions at higher frequencies, which leads to the over frequency with 20log(f1/f2) affecting the signal quality
need for accommodating many more transceivers into a small as well as some fundamental receiver aspects. The higher
space. level of phase noise requires higher margins and hence further
reduce the maximum output power.
In addition to the propagation conditions there are several
additional technology challenges arise at high frequencies, 5. The receiver noise figure, which directly relates to the uplink
for example: link budget. Hence, the uplink coverage also degrades over
frequency. As an example, the noise figure for bands ~3 GHz
1. The power amplifier output power drops by around 20 dB is typically ~5 dB better compared to bands around 30 GHz.
when the frequency increases a factor of 10, see figure. The In addition, there is a complex dependency between noise
figure shows the saturation point from the power amplifier figure, linearity of receiver and dynamic range of the receiver.
(PA) for a given semiconductor technology, e.g. CMOS.
6. At higher frequencies, the available bandwidths are larger.
Other technologies show a similar frequency dependencies.
Within the radio, digital algorithms such as PA linearization
Thus, maximum output power level will be lower at higher
become more challenging at larger bandwidths. Also,
frequencies. Therefore, many power amplifiers will be needed
larger volumes of data must be processed, which puts
to provide sufficient total output power from the Massive
more stringent requirements on interfaces. Data handling
MIMO radio.
is exacerbated due to the large number of transceivers for
2. The power efficiency degrades with increasing frequency, Massive MIMO radio.
see [1, Ch. 11].
Thus, the RF performance and efficiency degrades generally
3. In addition to reduced size, the degraded output power and with frequency. The listed issues 1-6 all contribute to degrading
efficiency make the thermal design quite complex at higher coverage and increases the requirements on the hardware
frequencies, which results in a delicate balance between solution. This makes mmWave Massive MIMO solutions more
linearity, available power and power efficiency. challenging compared to, for example, mid-band solutions, since
such solutions take a significant step up in frequency.
mmWave solutions for increasing total

antenna
mmWave gain
solutions for increasing No chan
total antenna gain

Need large antenna gain both
at network and UE side
Beam management - network
and UE need to track and agree
on best beams and paths
Large Many antenna elements

antenna gain and associated radio chains
Analog architecture with Beam
time-domain beamforming management
Large
bandwidth
One of the most important solutions to improve coverage is to To support the combination of large bandwidth and many
increase the antenna gain, both at the network side (the Massive antenna elements without too high complexity, analog
MIMO radio) and at the UE side. implementations of time-domain beamforming as described
in [TP6] have typically been adopted by the industry to strike
As a typical antenna element has an area proportional to the a relevant balance between performance and complexity. One
square of the wavelength, it is possible to use more elements and important limitation of time-domain beamforming is that it is
thereby increase the antenna gain without increasing the physical only possible to receive or transmit in one direction at a time.*
antenna size for a higher frequency as compared to a lower This observation together with the need to have high antenna
frequency [TP1, p. 9]. Comparing a mid-band frequency such as gain means that only one narrow beam can be transmitted
3 GHz with a high-band (mmWave) frequency such as 30 GHz, or received at a time. This limits the observability and makes
the number of elements will be 100 times larger for mmWave it challenging to determine in which direction to transmit or
as compared to mid-band for the same physical antenna area. receive. Typically, a multi-step approach must be used, in which
This means 20 dB higher antenna gain for mmWave without different directions are swept over time in order to find the best
increasing the size of the antenna. At the same time, however, the transmit/receive direction. Furthermore, features using very
antenna beams will become narrower as beamwidth is inversely narrow beams become more sensitive to various impairments
proportional to gain. Also, many antenna elements (even though and misalignments. The process of finding and maintaining
the elements are grouped into subarrays) give many associated matchings directions at both the network and UE side is
radio chains that add to hardware and software complexity. referred to as beam management for which support has been
standardized, [TP4, p. 61]. Note that allowing for more antenna
elements, hence increasing the array size, will result in even more
narrow beams with even higher gain. However, since the beams
are narrower, many more beams are needed to maintain the
same angular service area compared to a smaller array and more
beams needs to be searched which will increase control signaling
overhead and the time to find the best beam.
*) In an implementation with one time-domain beamformer

per polarization, it may be possible to transmit/receive in two
directions simultaneously, one on each polarization.
mmWave solutions for coverage,

capacity
mmWave and userfor
solutions throughput
coverage, N
capacity and user throughput
Coverage Extended range

• Link budget enhancements
• Frequency band interworking
Beam management
Capacity Cell shapes
• Large bandwidth Coordinated

• Spectral efficiency enhancements multi-layer network
User throughput and latency

• Large bandwidth Frequency division multiplexing
• Short TTI
As discussed in the previous page, the need to support very large mmWave has the potential to provide substantial capacity, very
bandwidths and high antenna gains, both at network and UE high user throughput and low latency due to the large amount of
side, has steered the industry to choose an analog architecture available bandwidth. There are, however, two main challenges
with time-domain beamforming for current mmWave products. for mmWave to reach the full potential. The bandwidth must
This is different compared to TDD mid-band Massive MIMO be made available to the users, that is, coverage needs to be
products that use a digital architecture and frequency-domain sufficiently good, and the analog architecture with time-domain
beamforming. This impacts some of the features at mmWave beamforming restricts the possibility to multiplex different
and requires some new 3GPP standard support. signals and/or users, both spatially and in frequency, and
enforces the use of beam management that adds access latency
Coverage can be improved by considering radio and antenna- and consumes capacity.
related attributes like output power, sensitivity and antenna gain,
both at network and UE side. System attributes like bandwidth This chapter will describe some selected features for mmWave
and slot format, and frequency interworking where coverage- that are important for improving coverage, capacity and
limited signals or channels are sent on a lower frequency band user throughput. While some features are tailored towards
with better coverage can also be considered. Output power and a particular performance aspect, e.g. coverage, most of the
receiver sensitivity are parameters predominantly addressed features impact all areas. This is illustrated in the right part of
in the hardware domain, as indicated on page [TP8, p. 127]. the figure. In addition to these features, the features used for
Increasing the antenna gain is central for mmWave and as mid-band [TP7] can in general be used also for mmWave.
discussed in the previous page it is achieved by increasing the
number of antenna elements both in the Massive MIMO radio
and the UE. See [1, Ch. 7] for further details. Also, there are
several RAN software features that are designed to improve
coverage as well as other performance aspects.
Beam management for coverage

at mmWave
Beam withfor
management analog beamforming
coverage at No chang
mmWave with analog beamforming

SSB1
Step 1: SSB2
Wide beam sweeping
Reduced beam sweeping latency SSB3
and system overhead for initial access
SSB4
Step 2:
CSI-RS1
Narrow beam sweeping CSI-RS2
Maximized beamforming gain to CSI-RS3
reach users with gigabit speed
Beam management to create and maintain best beam pair between gNB and UE
As outlined in the introduction above, beamforming is a key It utilizes wide beams for initial access and narrow beams for
component of mmWave solutions for maximizing the end-to- data. Initially, at connection setup, a set of wide beams (e.g.
end antenna gain to improve coverage. For mmWave, both using different SSBs) are swept through the service area. The UE
the base station and the user equipment (UE) need to use measures the SSB quality to determine the best wide beam to
transmit/receive beamforming with highly directional beams establish the connection. In a second step, a set of narrow beams
to maximize coverage. Due to the analog architecture with mapping to the selected wide beam coverage area are swept with
time-domain beamforming, this means that a beam pair needs measurement reference signals (e.g., CSI-RS), so that the UE can
to be created and maintained between the base station and the measure and report on the preferred narrow beam. Thus, the base
UE. The accuracy of the selection, and the delay in obtaining and station can refine the beam selection to the narrow beam with
updating the beam pair while the UE is moving, affect end-to-end highest beamforming gain to transmit data.
performance and the quality of service. This process is referred
to as beam management and an effective and efficient beam The two-level beam management, as used by Ericsson, enables
management framework plays a significant role in mmWave a high antenna gain and reach many users with 5G gigabit
deployments. speeds with multiple narrow traffic beams, while maintaining low
overheads and reducing initial access latency with wide access
There are several ways to do beam management based on beams. It also provides instantaneous beam switching with zero
hardware realization/capability and with the toolbox provided transmission interruption, so that users may enjoy smooth data
by 3GPP [TP4, p. 61]. One example is the two-step beam streaming when moving frequently across narrow beams within
management procedure shown in the figure above. a wide beam.
Different cell shapes are suitable

for different
Different deployment
cell shapes scenarios
are suitable for N
different deployment scenarios

Hotspot – example 1 Hotspot – example 2 Macro – example 1
12 wide beams 24 wide beams 12 wide beams
Service area: 120°x 30° Service area: 120°x 30° Service area: 120°x 15°
15 deg
30 deg
30 deg
-60º 0º 60º
-60º 0º 60º
-60º 0º 60º
Wide beam Narrow beam Wide beam Narrow beam Wide beam Narrow beam
A coverage shape defines the azimuth (horizontal) and elevation In other words, to achieve larger initial access coverage, wide
(vertical) coverage of the service area. A properly selected beams with higher beamforming gain are needed. If the target
coverage shape ensures that all UEs in the targeted service area service area can be smaller, which is often the case for rooftop
are covered well by the envelope of the wide beams carrying cell- deployments where less coverage area in elevation is needed, it
defining signals such as SSB in NR. There are various coverage is possible to adopt higher gain wide beams meanwhile fewer
shapes suitable for different deployment scenarios. For example, wide beams is needed.
hotspot shapes are suitable for street deployments where a wide
area in both azimuth and elevation shall be covered. In a typical Many wide beams implies more overhead for broadcasting
3-sector cell, a hotspot shape normally covers a service area of cell-specific information as well as more overhead for beam
±60° in azimuth and ±15° in elevation. Meanwhile, macro shapes management. Therefore, there is a trade-off between initial
are suitable for e.g. rooftop deployments where less coverage access coverage and overhead caused by too many wide beams.
area in elevation is needed. Data channel coverage, or app coverage, is dependent on the
beamforming gain of narrow beams. Therefore, a two-level
A coverage shape also defines the required gain of the wide beamforming approach enables the possibility to choose
beam. A wide beam or a set of wide beams that cover an entire an appropriate number and shape of wide beams, whereas
target service area, are utilized for initial access (broadcasting maximum beamforming gain on the narrow beam can always
SSBs and other system information etc.). Coverage of initial be achieved by creating many narrower beams independent
access ideally should be same as coverage of data channel. of number of wide beams. With this, the system can achieve
The signaling transmitted in initial access is carried in physical sufficient access coverage and excellent app coverage, while
channels adapting lower modulation and coding scheme. keeping the system overhead as low as possible.
Therefore, the required signal-to-noise ratio (SNR) used in initial
access is often lower than required SNR for data transmission. Ericsson provides adequate predefined cell shapes targeting
That is the reason why wide beams are adopted for initial access various typical deployment scenarios. With the help of AI/ML
to achieve moderately high beamforming gain. techniques, cell shapes, or beam sets, can be selected or tuned
depending on the actual spatial UE distribution in particular
For a given antenna array, higher beam gain means narrower deployments.
beam. The wide beams illustrated in Hotspot – example 2
are half of the beam width in azimuth compared with the
wide beams in Hotspot – example 1. The beamforming gain
in example 2 is therefore approximately twice as high as in
example 1. This in turn leads to that more beams are needed in
order to cover the same service area in azimuth and elevation.
Coordinated multi-layer network to

maximize
Coordinatedthe network
multi-layer performance
network to maximize Make su
Notes pa
the network performance
Carrier aggregation Dual connectivity
mmWave
Mid-band
Low-band
Coverage area A Coverage area B Coverage area C Coverage area D

All bands have Mid-band and low-band mmWave and low-band Only low-band
coverage have coverage have coverage has coverage
Coordinated communication using all the available frequency Low-bands offer excellent coverage due to the inherent physical
bands maximizes the benefits of each of the bands where propagation properties on low-frequency bands and the
best needed. Efficient coordination between the different use of FDD. Low bands are therefore the most economically
frequency layers in the network is very important to reach feasible way to enable nationwide 5G coverage. The available
a superior network experience, i.e. coverage everywhere, bandwidth on the low bands is, however, limited and thus the
substantial capacity almost everywhere and extreme capacity offered capacity is limited. Mid-band spectrum is the sweet spot
where needed. This is particularly important for mmWave for offering a 5G experience, since the coverage is good, the
deployments as the performance potential is substantial, but available bandwidth is substantial and the conditions for using
coverage can be challenging. The 3GPP 5G standard provides an Massive MIMO are excellent. Thus, using mid-band with Massive
extensive toolbox to achieve efficient coordination with solution MIMO provides good coverage and substantial capacity and user
components such as carrier aggregation, cross-carrier scheduling throughput. Mid-bands are also globally available, thus enabling
and dual connectivity, see [TP4, p. 47]. economy of scale. The available bandwidth on mmWave
is very high and hence mmWave has the potential to offer
unprecedented peak rates, low latency, and high capacity. The
mmWave spectrum is a valuable resource when targeting sports
arenas, densely populated urban areas with a large number of
mobile devices, and time-critical services. Beamforming, beam
management, traffic management and efficient multi-layer
interworking enable a fully coordinated multi-layer network can
provide the best performance and flexibility to secure service
differentiation.
Extended range for fixed wireless access

Extended range for fixed wireless access
Covered by mmWave Covered by mid-band,

and mid-band. but not by mmWave (no line-of-sight).
Offer wireless fiber

service with mmWave
capacity
Serve more
households with mmWave base range mmWave extended range – several kilometers
extended range
Despite the challenging propagation on mmWave, gigabit The larger the distance is from a household to the radio site, the
performance up to several kilometers using mmWave for FWA more challenging it is to achieve good line-of-sight conditions.
has been proven in field. This requires the use of high-power The combined use of mmWave and mid-band provides an
radios, high-power CPEs, and beamforming techniques with efficient network solution, where mmWave serves household
RAN software innovation to extend the mmWave cell range in good conditions and mid-band serves households in less
and accommodate the increased propagation time at longer favorable conditions. With the capacity offload enabled by
distances. mmWave, mid-band can serve more households at more distant
and challenging locations. This provides an opportunity to offer
The 3GPP NR specification allows for large cell ranges high-end wireless fiber services in more rural areas, serve more
on mmWave. The guard period – the gap between the downlink households per radio site, and support higher data consumption,
and uplink in the TDD configuration – caters for the transmission which makes FWA based on mmWave very efficient and a highly
roundtrip time and the time needed for the equipment to switch attractive use of mmWave spectrum.
from reception to transmission. An extended cell range is enabled
using a larger guard period. The drawback is slightly reduced
downlink peak rates for all users in the cell. FWA extended
range can be used to serve selected households in good
signal conditions at several kilometers distance.
Frequency division multiplexing (FDM)

to improve
Frequency capacity
division multiplexing (FDM) to The das
time freq
improve capacity
Without FDM With FDM
ncy
cy
ue
n
eq
…
ue
Fr
eq
Fr
Time
Time
Frequency division multiplexing (FDM) is a technique in order to Frequency division multiplexing (FDM) improves the spectrum
serve multiple UEs in the same timeslot, where UEs are separated utilization by creating multiple simultaneous beams covering
in frequency. different parts of the frequency band in the same time slot,
meaning that more UEs can be served simultaneously. For an
One analog time-domain beamformer creates one beam for one analog architecture with time-domain beamforming, FDM
timeslot covering the entire instantaneous bandwidth, typically operation can be achieved by having multiple beamformers
consisting of several frequency carriers. This makes it possible to operating on different parts of the frequency range, see [TP6, p.
offer extremely high data rate to a single user. However, when the 84] for further details.
network is loaded with not only UEs with larger data packets but
also UEs with smaller packets, it becomes less efficient as some of In brief, FDM can improve spectrum utilization for a loaded
the frequency carriers might not be able to be utilized. The small network. This in turn improves system throughput, ensures more
packets can be either payload or control signaling. connected users, and reduces latency as more users can be served
simultaneously instead of waiting for their own timeslot.
Summary
Summary N
● mmWave can provide very high capacity

due to the large available bandwidth
● mmWave coverage is limited due to
higher frequency and deployment shall
focus on targeted high traffic areas
● Architecture and hardware solutions are
designed to provide a well-performed
and cost-efficient solution
● Software solutions can further improve
coverage and capacity
mmWave solutions can provide very high capacity and user The most important use of mmWave in the near-to-mid term
throughput, since the available bandwidths are very large. is to enhance the total network performance with respect
This makes mmWave solutions attractive options for capacity to capacity and end-user throughput via inter-working with
expansion and enhancement of user experience. lower frequency bands, e.g. use mid-band when the coverage
on the mmWave frequency bands is insufficient. Additionally,
The main challenge at mmWave frequencies is that the mmWave can also be used to provide very high performance in
propagation conditions are more challenging at higher hot spots, particularly in areas with pre-dominantly line-of-sight
frequencies and hence coverage cannot reach the same level propagation conditions or when the propagation path gain is
as for mid-band. Due to this, the preferred deployments are dominated by reflections.
targeted high traffic areas, e.g. sports stadiums, town squares,
etc. Some other deployments are also well suited for mmWave.
For FWA, where the UE can be designed with higher antenna
A number of solutions are made that improve coverage gain and output power and can be located outside the building
substantially. For example, beamforming is used in both the to avoid penetration losses, mmWave can also be useful.
Massive MIMO radio and in the UE, analog beamforming In environments dominated by diffraction and/or building
is generally used in the industry get a balance between penetration the challenges largely remain.
performance and complexity. Coverage on mmWave bands
does however remain a challenge. Additional improvements on By applying solutions in software and hardware and utilizing the
mmWave solution may further reduce the coverage difference 3GPP standard support, very high capacity and user throughput
compared to mid-band. The higher losses at the mmWave bands as well as substantial coverage improvements can be achieved.
related to diffraction and building penetration remain.
The hardware solutions have improved substantially since the
early deployments and continue to improve in the years to come.
9. Network performance
9. Network Performance
Ericsson | The Massive MIMO Handbook — Technology Primer 9. Network Performance 137
Network performance
—Outline
Network performance – Outline N
C
• Radio network performance evaluations
• Key aspects impacting Massive MIMO

performance
• Capacity dimensioning
• Deep dives
• TCO analysis
The radio network performance depends on many aspects that This chapter will give some insights into the area of radio
interact in an intricate manner, and it is important to understand network performance for Massive MIMO. Predictions based on
the driving mechanisms and have models that capture the reality simulations will also be compared with measurements in real
sufficiently well while still being computationally efficient. networks. The chapter is divided into four parts:
The most important tool for quantitative performance
evaluations is advanced network simulations which capture the • Very short introduction to radio network evaluations
essential parameters that affect performance. Measurements in
• An overview of key aspects impacting Massive MIMO
testbeds and in real networks are also important for verification
performance will be given, followed by some illustrations
and demonstration but cannot support large-scale quantitative
of capacity dimensioning with TDD mid-band and Massive
evaluations under known conditions. Ericsson is at the forefront
MIMO
in the area of radio network performance evaluations using
state-of-the-art knowledge and models. The predictions based • Deep-dives into a few selected areas including the impact of
on simulations have proven to match real network performance the number of MU-MIMO layers, Interference Sensing, benefits
on countless occasions. of frequency band interworking, impact of type of channel state
acquisition approach (codebook-based vs reciprocity-based),
gains of larger antennas and mmWave performance
• Total cost of ownership (TCO) analysis for mid-band

deployments
Performance evaluations can guide service providers when it

comes to network evolutions and product needs. They are also
vital in the research and development (R&D) flow. It provides
an understanding of how different deployments, configurations
and algorithms perform before making costly R&D product
investments.
Radio network performance evaluations

Radio network performance evaluations
Network characteristics matter! Key performance metrics

Need to capture a diverse and complex reality Coverage, capacity and end-user throughput
Cell-edge user
throughput
System More cap
able netw
Antenna ork
assumptions User through
characteristics gain
Baselin
Algorithm e network
assumptions
Radio
propagation
User Network
throughput capacity gain
Deployment … requirement
characteristics
Baseline Served traffic

capacity
Performance on radio network level depends on many aspects Some key performance metrics reflecting needs in the network
that interact in an intricate manner, and it is important that there are coverage, capacity, and user throughput [Ch. 1, p. 14]:
are simulator models that capture the essential characteristics
of the reality and of the radio access network features to be • Different metrics reflecting coverage can be envisioned, but
evaluated. Depending on the feature to evaluate and the results they all try to quantify how reliably user data and control
of interest, certain aspects are often more important than others, information can be received. A common measure of coverage
and for Massive MIMO features, spatial characteristics are is to look at the achievable user throughput for the worst users
obviously very important. Some key aspects impacting Massive in the network, the so-called cell-edge users (often defined as
MIMO performance include deployment environment, antenna the 5th percentile worst users in the network). Coverage is a key
characteristics and algorithm assumptions. These aspects will performance metric when dimensioning a network as it relates
be further described in the following slides. System assumptions, to the number of required sites and therefore also directly
such as frequency band, output power and bandwidths can also affects the total network roll-out cost
have a significant impact on the performance.
• For packet-based MBB services, capacity is often expressed
as the average served traffic (bits per second) per cell or per
area unit given a certain resource utilization or user throughput
requirement at a specific UE percentile, often the cell-edge user
percentile. Capacity is a key metric to decide how many users
or how much traffic the network can sustain with maintained
coverage. The capacity is often of high interest for operators as
it reflects the production cost (cost per bit)
• User throughput reflects the user-experience and is

closely related to coverage and capacity. Throughput will
differ between users and depends on their respective link
quality, total cell or network traffic load, scheduling, system
assumptions, etc. It is common to assess user throughput gains
for different user percentiles, for example, the 5th (cell-edge),
50th and 95th percentiles, at specific network loads
9.1 Key aspects impacting

Massive MIMO
performance
Key aspects impacting Massive

MIMO performance
Key aspects impacting Massive MIMO N
performance
W
Antenna array
Antenna configuration 4x1 !" 2x8
Subarray 4x1 !"

Array of subarrays, 2x8
Dual-polarized
antenna element pair
+
Two radio chains per subarray; one per polarization 2x2𝑥𝑥8 = 32 radio chains in total
Deployment scenario Algorithm choice

Building
height
Dense urban Urban Suburban Rural

ISD 100-500m 600-1000m 1500-4000m ~5-20km Classical beamforming Generalized beamforming
This slide highlights three factors that have a substantial impact The deployment scenario (lower left figure) including inter
on the performance of Massive MIMO evaluations: the antenna site distance, the radio environment, user distribution in
configuration [Ch. 1, p. 17], the deployment scenario [Ch. 2, p. both horizontal and vertical domains, etc. has a substantial
30-31], and the Massive MIMO algorithm choice [Ch. 1, p. 15], impact on the performance and how different Massive MIMO
[Ch. 3.3]. configurations perform. In particular, the user distribution in the
vertical domain impacts what subarray size that is feasible and
The antenna configuration (upper figure) is central for the the usefulness of vertical domain beamforming, and the inter site
performance. The antenna area determines the maximum distance affects the coverage.
antenna gain (coverage). The subarray consists of a number
of semi-statically combined dual-polarized antenna element The Massive MIMO algorithm choice (lower right figure) is yet
pairs and is the smallest dynamically controllable entity (two another important factor affecting the performance. Two main
radio chains per subarray, one per polarization). It determines classes of beamforming are so-called classical beamforming and
the angular coverage area (in the vertical domain), i.e. the area generalized beamforming [TP3, p. 27]. Two other central aspects
where beamforming can be done without significant gain drop. are how the channel state information needed for beamforming
A small subarray (few vertical elements) gives a wider coverage is acquired and what granularity of channel state information
area but also less gain, whereas a larger subarray (more that is needed. See also [TP3, p. 37-38], [TP4, p. 49, 51-53] and
elements) gives a narrower coverage area but also higher gain. [TP7, p. 101-103].
The number of radio chains (Tx and Rs) determines the degrees
of freedom (or steerability) for beamforming. Stacking subarrays
horizontally (radio chains in the horizontal domain) provides
horizontal domain beamforming and stacking subarrays
vertically facilitates vertical domain beamforming. Stacking
subarrays of an appropriate size (increasing the number of radio
chains) enables the use of a larger antenna, hence more gain
that can be steered over the entire service area.
5G networks offer a rich toolbox

for improving coverage
5G networks offer a rich toolbox Blue cyl
for improving coverage

Bandwidth
Key components for

Beamforming
improving coverage
High-band
• Beamforming
• Frequency interworking
Mid-band Beamforming
Low-band
Coverage
Coverage is a key performance metric when dimensioning a • Beamforming is a key feature for increasing coverage.
network as it relates to the number of required sites and therefore Maximum antenna gain is determined by the antenna size,
also directly affects the total network roll-out cost. Coverage hence from a coverage perspective, both uplink and downlink, it
matters from day one when deploying a new 5G network, is beneficial to have as large antenna as possible. One potential
whereas capacity needs grow gradually. challenge with a large antenna is that the beamwidth is
inversely proportional to the gain. Hence, if you want gain, the
Most multi-antenna techniques are essentially frequency corresponding beams will be narrower. One solution to this is
agnostic. However, as discussed in [TP1, p. 9], the frequency to employ UE-specific beamforming, where subarrays of a size
has a profound impact on propagation properties and antenna that fits the deployment scenario are stacked horizontally and/
characteristics affecting coverage using the existing site grid, or vertically. Stacking more subarrays renders a larger antenna,
hence impacting the design and systemization of Massive MIMO. hence more gain, and by using UE-specific beamforming, the
Examples of propagation effects that depend on frequency gain can be directed over the entire coverage area given by the
include foliage, diffraction, wall loss, etc. subarray size.
Two key network components for improving coverage are • Frequency interworking: It is important that network
frequency interworking and beamforming: algorithms explore that different frequencies exhibit different
coverage characteristics. By using carrier aggregation with
cross carrier scheduling, one can maximize the area where
downlink data can be sent on mid-band by sending coverage
limited, but essential data and control info (often uplink
related) on a lower band with better coverage.
Other solutions to the coverage problem include (see [TP7.1]

for details): dynamically adapt beamforming algorithms to
the coverage situation, e.g. use more advanced beamforming
algorithms that rely on better channel state information in areas
with good coverage, and more basic beamforming algorithms
requiring less accurate channel state information when coverage
is worse. Increased output power (including power boosting),
better receiver sensitivity, more robust coding or repetition are
examples of other common ways of increasing the coverage.
Capacity simulations for Massive MIMO

Capacity simulations for Massive MIMO N
W
Different deployment scenarios call for different antenna configurations
7
same antenna area
DL capacity gain relative to 2T
5
2T 4T 8T 16T 32T 64T
𝟏𝟏𝟏𝟏𝐱𝐱𝐱𝐱 𝟏𝟏𝐱𝐱𝟏𝟏 𝟏𝟏𝟏𝟏𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟏𝟏𝐱𝐱𝟐𝟐 𝟖𝟖𝐱𝐱𝐱𝐱 𝟏𝟏𝐱𝐱𝟒𝟒 𝟖𝟖𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟏𝟏𝐱𝐱𝟖𝟖 𝟒𝟒𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟐𝟐𝐱𝐱𝟖𝟖 𝟐𝟐𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟒𝟒𝐱𝐱𝟖𝟖
4
𝐒𝐒𝐒𝐒 𝐒𝐒𝐒𝐒
Adding enhanced beamforming in the vertical domain

3
2 Enhanced beamforming in the horizontal domain
0
2T 4T 8T 16T 32T 64T
suburban/rural urban low/mid rise dense urban high rise
Capacity is a key metric to decide how many users or how much In the rural/suburban deployment (~1500 m inter-site distance),
traffic the network can sustain with maintained coverage. the gains of going from 16T to 32T or 64T are rather small.
Capacity is here defined as traffic per area unit (bits per second The main reason is that the 8x1 subarray used in 16T fits well
per area unit) given a user throughput requirement of 10 Mbps the narrow angular spread of users in the vertical domain.
for the 5th user percentile (so-called cell-edge users).
In the urban low/mid-rise deployment (~500 m inter-site
The figure shows capacity gain relative to 2T classical distance), the gains of going from 16T to 32T is significant,
beamforming for different antenna configurations in different but the additional gain by going from 32T to 64T is small.
deployments, as also discussed in [TP2, p. 21]. The antenna
configurations range from 2T to 64T, and generalized Results in the dense urban high-rise deployment stand out with
beamforming using a mixture of codebook-based and substantial gains by going from 16T to 32T to 64T. Compared
reciprocity-based channel state information is used. to the other deployments, this scenario has smaller inter-site
distance and tall high-rise buildings, hence better coverage,
The key takeaway is that the effectiveness of a specific antenna substantially more spread of users in the vertical domain and
configuration is deployment dependent. More specifically, more channel angular spread. Also, the inter-cell interference
horizontal domain beamforming (2T to 16T) is useful in all becomes problematic at high load. All this combined makes
deployments, and a common rule-of-thumb is that capacity Massive MIMO superior to other solutions.
increases by 30-50% by doubling both the antenna area and
the number of radio chains in the horizontal domain, e.g. going
from 2T to 4T or from 8T to 16T. Gains of vertical domain
beamforming (16T to 64T) depends on the deployment, and a
key parameter is the angular distribution of the users in relation
to the size of the subarrays
Capacity simulations for Massive MIMO

Capacity simulations for Massive MIMO No change
Wrong font in
3
Suburban/rural
Capacity gain relative to 4T
Urban low/mid-rise
~50%
2,5
4T 16T 32T 64T
𝟏𝟏𝟏𝟏𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟏𝟏𝐱𝐱𝟐𝟐 𝟖𝟖𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟏𝟏𝐱𝐱𝟖𝟖 𝟒𝟒𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟐𝟐𝐱𝐱𝟖𝟖 𝟐𝟐𝐱𝐱𝐱𝐱 𝐒𝐒𝐒𝐒 𝟒𝟒𝐱𝐱𝟖𝟖
2
~10%
1,5
1
16T 32T 64T
Classic beamforming Generalized beamforming
The previous slide illustrated that the performance gains of There is a deployment and antenna configuration dependency
deploying Massive MIMO depends on the deployment scenario. on the gain of generalized beamforming over classical
The results in this slides illustrates the performance impact of beamforming. For example, for the 16T antenna configuration,
different beamforming algorithms. the gain of generalized beamforming compared to classical
beamforming is similar for both deployment scenarios, whereas
The figure shows capacity gains relative to 4T classical the corresponding gain for the 64T antenna configuration is
beamforming for different Massive MIMO antenna ~10% and ~50% in suburban/rural and urban low/mid-rise
configurations using classical or generalized beamforming [TP3, deployment scenarios, respectively.
p. 27] in rural/suburban and urban low/mid-rise deployment
scenarios. Classical beamforming uses SU-MIMO codebook-
based channel state information with a grid-of-beams type of
codebook, whereas generalized beamforming uses a mixture
of SU/MU-MIMO and codebook-based and reciprocity-based
channel state information.
Capacity results from initial trials

Capacity
of results
Massive MIMOfrom initial trials N
A
of Massive MIMO
Results match our
state-of-the-art radio LTE 64TR Massive MIMO TDD
network predictions
4
Capacity gain vs. 2T2R
Early trial and deployment result from LTE deployments of

AIR6468 64T64R radio in different networks around the world,
represented by different bars in the figure.
The observed results of ~2-3x capacity of 64T64R compared to

2T2R in different networks coincide well with the predictions
made using simulations, see e.g. [TP9, p. 157].
9.2 Illustration of capacity

dimensioning for
TDD mid-band
Quantifying the benefit with mid-band

TDD and Massive MIMO
Quantifying the benefit with mid-band N
TDD and Massive MIMO

Network evolution scenarios
1 2 3
FDD+TDD FDD+TDD with

FDD only
with 8T8R Massive MIMO
Sub-1 GHz FDD, 20 MHz Sub-1 GHz FDD, 20 MHz Sub-1 GHz FDD, 20 MHz Study area:
2 GHz FDD, 40 MHz 2 GHz FDD, 40 MHz 2 GHz FDD, 40 MHz Atlanta midtown, north of downtown
Size 2 km2 (1550 x 1250 m)
3.5 GHz TDD, 100 MHz 3.5 GHz TDD, 100 MHz
Urban scenario
The following slides will illustrate how different radio The main takeaways from this example are
deployment options will handle the expected traffic growth over
the coming years. • Adding NR TDD mid-band with Massive MIMO provides a
future proof investment that can handle the anticipated traffic
The deployment scenario is Atlanta midtown (US dense urban growth for many years
scenario), and results are based on multi-band simulations for
three different radio deployment options: • Low-band frequency deployments are essential to ensure
uplink coverage and superior uplink user experience
1. FDD bands only: FDD 700 MHz with 20 MHz bandwidth
and 2T2R radio, and FDD 2 GHz with 40 MHz bandwidth
and 4T4R radio
2. FDD bands plus 3.5 GHz TDD with 100 MHz and 8T8R radios
3. FDD bands plus 3.5 GHz TDD with 100 MHz and Massive
MIMO (64T64R)
The coming slides will show user throughput (represented
by different colors) for the individual users in the city map at
different traffic loads. The first four slides present downlink
results, followed by three slides showing uplink results
Downlink user experience visualized

Downlink
Downlink userUser Experience
experience Visualized
visualized No chan
Mbps
Baseline load: NR TDD provides massive boost in user experience
1 2 Traffic demand
[Gbps/km2]
750
3
year
500
2
250 FDD bands FDD + NR TDD 8T8R
100 2025
3
1
Low network load: 2023
50
Baseline traffic demand
2021
of 0.2 Gbps/km2 (~2019)
0.2 .2
10
0
This slide shows downlink user throughput (represented by It is also seen that differences between the 8T8R radio
different colors) for the individual users in the Atlanta city map deployment and Massive MIMO are only noticeable in a few
at a baseline traffic load of 0.2 Gbps/km2, corresponding to the regions marked with blue circles. These areas contain high-loss
traffic demand ~2019. buildings with high path loss, hence the extra beamforming gain
offered by Massive MIMO is beneficial.
Results for the three different radio deployment options
described in the previous slide are marked with 1, 2 and 3,
respectively.
It is clear from the results that adding NR TDD mid-band to

existing FDD bands provides a substantial boost in performance.
Only FDD bands (1) provides downlink speeds in the range of
100-250 Mbps but by adding TDD mid-band, speeds exceed
500 Mbps.

Downlink user experience visualized N
x2 load: FDD bands become gradually over-loaded

Mbps
1 2 2 Traffic demand
[Gbps/km2]
750 3 year
500
2
2025
100 3 3
1
x2 increased network load: 2023
50 Traffic demand of
0.4 2021
0.4 Gbps/km2 (2021)
10
0
This slide shows results after doubling the baseline traffic By adding the TDD band, the user experience is mainly on green
demand, now reaching 0.4 Gbps/km2, corresponding to the levels throughout the network, but compared to the previous
traffic demand in ~2021. slide, some more yellow areas are appearing, in particular when
employing 8T8R.
The FDD bands alone are not sufficient to deliver expected
downlink speeds. There are many orange and red areas in the
city model indicating downlink speeds that do not meet the
expectations from a 5G experience.

Downlink user experience visualized No chan
x4 load: emerging benefits of Massive MIMO vs conventional 8T8R

Mbps
Traffic demand
1 2 2 [Gbps/km2]
750 3
year
500
2
100
2025
3 3
1
0.8
50
Traffic demand of 2021
0.8 Gbps/km2 (>2023)
10
0
This slide shows results after yet another doubling of the traffic
demand, now reaching 0.8 Gbps/km2, corresponding to an
expected traffic demand in 2023+
Clearly, only using the FDD bands cannot support the traffic
demand, and adding mid-band becomes necessary. It is also
clear that Massive MIMO starts to provide substantial gains
compared to 8T8R. Massive MIMO provides downlink user
throughputs in the order of 250 Mbps in the network while for the
8T8R radio solution, there are many users experience 100 Mbps
or lower.

Downlink user experience visualized N
x8 load: substantial gains of Massive MIMO over conventional 8T8R

Mbps
Traffic demand
2 [Gbps/km2]
1 2
750 3 year
500
2
1.6
2025
100
3
3 1
50
high traffic demand of 2021
1.6 Gbps/km2 (>2025)
10 0
As a final illustration, results with yet another doubling of the

traffic demand, reaching 1.6 Gbps/Km2 corresponding to the
expected traffic demand in 2025+, are shown.
These results clearly illustrate the superior performance offered

by adding TDD mid-band with Massive MIMO to the existing
site solution. The radio solution with FDD bands plus mid-band
with conventional 8T8R starts to become over-loaded, while the
corresponding radio solution using Massive MIMO on mid-band
still provides a superior user-experience with downlink speeds
well above 250 Mbps.
Uplink user experience visualized

Uplink user experience visualized No chan
Baseline load: lower FDD bands are key for uplink coverage
Mbps
Traffic demand
1 2 [Gbps/km2]
50 0.6 year
25
0.4
5
2025
3 0.2
Low network load: 2023
1
Baseline traffic demand of 0.08 2021
0.08 Gbps/km2 (~2021)
0.1 0
To complement the downlink results shown in the previous These results illustrate that low-band deployments are essential
slides, the following slides will show uplink user throughput. to reach superior uplink performance. One main reason being
The baseline traffic load is 0.08 Gbps/km2, corresponding to that uplink coverage is challenging on higher frequency bands,
the traffic demand ~2021. and it is to a large extent the users with worst coverage that
dictate the overall network performance as those users consume
Results for the three different radio deployment options a lot of the radio network resources. Letting a low-band with
described in [TP9, p. 147] are marked with 1, 2 and 3, better coverage serving the worst users, will not only help
respectively. the coverage-limited users, but also boost the entire network
performance.
The results show only small differences in user experience for
the three different radio solutions. As opposed to the downlink Further discussions of the benefits with frequency band
results, the addition of mid-band does not bring any significant interworking and carrier aggregation with cross-carrier
performance gains compared to using only FDD bands. There scheduling can also be found in [TP7, p. 107] and [TP9, p.
are, however, some users that get a better experience with the 159-160] .
addition of mid-band, especially with Massive MIMO.
Uplink user experience visualized

Uplink user experience visualized 20
x4 load: modest throughput gains by NR TDD

Mbps
Traffic demand
1 2 [Gbps/km2]
50 0.6 year
25
0.4
0.28 2025
5
3 0.2
1
medium traffic demand of 2021
0.28 Gbps/km2 (~2025)
0.1 0
This slide shows results after quadrupling the baseline traffic

demand, now reaching 0.28 Gbps/km2, corresponding to the
traffic demand in ~2025.
The results are still rather similar for all three radio deployment
solutions, although using FDD bands only shows more yellow
regions than FDD bands together with mid-band, and FDD plus
mid-band with Massive MIMO is slightly greener than the other
two solutions.
Uplink user experience

visualized
Uplink user experience visualized Remove
x8 load: Massive MIMO provides

Mbps more consistent performance at high traffic
Traffic demand
1 2 [Gbps/km2]
50 0.6 0.6
year
25
0.4
5 2025
3 0.2
x8 increased network load : 2023
1
high traffic demand of 2021
0.6 Gbps/km2
0.1
0
This slide shows results with eight times the baseline traffic, now The main reason why the addition of mid-band, and especially
reaching a high traffic demand of 0.6 Gbps/km2. mid-band with Massive MIMO, boosts the performance
compared to using only FDD bands in this case is that the
This extreme traffic demand cannot be supported by using FDD mid-band can off-load the valuable low-bands. The capacity
bands only and many users get very poor user experience. The of the low-bands is not enough to support the traffic demand.
addition of mid-band to the FDD bands makes the situation The mid-band can serve many of the users with good experience,
much better with most users reaching above, or even well above allowing the low-bands to focus more on serving the users with
5 Mbps. It is also seen that mid-band with Massive MIMO poor coverage.
provides a much more consistent user experience in the network,
that is, less red or yellow areas compared to mid-band with
8T8R.
9.3 Deep dives
TDD mid-band
TDD mid-band MU-MIMO
MU-MIMO and and number No
number
of offrom
layers layerssimulations
from simulations
W
Opportunistic gains from MU-MIMO – rather few layers are typically needed
How many layers are typically scheduled? Capacity gains of allowing more layers
No restriction on max number of layers Restricting max number of layers
MU-MIMO friendly dense Urban high-rise deployment
Chapter TP 3 gave some insights into realistic expectations on Some observations are as follows:
MU-MIMO, that will be further illustrated here.
• The left figure shows the distribution of the number of
Downlink results for two MIMO favorable deployment scenarios scheduled layers as seen from the schedulers without any
are shown; dense urban high-rise and dense urban low/ restriction on max number of layers. It is seen that 2 layers
mid-rise. Both deployments have small inter-site distances clearly dominates, and it is rare to schedule more than 6 layers.
and large channel and user spread. The setup is TDD 3.5 GHz Slightly more layers are beneficial in the dense urban high-rise
with 100 MHz bandwidth, 200 W output power and 64T 8x8 deployment (no indoor systems are assumed, i.e. all indoor
antenna array using generalized beamforming. Furthermore, it is users are served by outdoor Macros)
assumed that each UE can have max two SU-MIMO layers. As an
example, when four layers are scheduled, it can be one layer to • The right figure shows the capacity gain relative to max four
four different UEs, two layers to two different UEs, or one layer to layers for different restrictions on the maximum number of
two different UEs and two layers to one UE. layers in the MU-MIMO friendly high-rise deployment. It is seen
that max 16 layers provides a capacity gain of 25%, but most of
the gain is achieved already with max 6 or 8 layers
Ericsson Interference Sensing substantially

improves performance in dense and highly
Ericsson networks
loaded Interference Sensing substantially improves Title cha
Remove
performance in dense and highly loaded networks
Downlink
performance Network
Good
+40% capacity gain
Poor
Interference sensing: Off Interference sensing: On
As discussed in [Ch. 1, p. 15], there are three main benefits of For special use cases or when the traffic load increases in
Massive MIMO: (1) beamforming that increases the signal denser deployments, it is reasonable to believe that inter-cell
quality, (2) spatial multiplexing that increases data rates, interference will become a severe problem that needs to be
and (3) nullforming that reduces interference. While the first handled.
two features are vital for a high performing Massive MIMO The Ericsson Interference Sensing feature provides an optimal
system, the benefits of the last feature, nullforming, are more trade-off between the three main benefits of Massive MIMO
discussed, at least when considering inter-cell interference. mentioned above. Inter-cell interference avoidance is an
There are several reasons for this. First, inter-cell interference intricate part of the feature, and the novel solution overcomes
suppression/avoidance by means of nullforming often provides the problems discussed above. Advanced signal processing
less gains than beamforming and spatial multiplexing in many algorithms solve the problem of acquiring accurate channel
deployment scenarios. Furthermore, there exists already other state information in an innovative manner, and the Ericsson
means for handling inter-cell interference like antenna tilt. Silicon manages the increased signal processing computational
Second, nullforming is often more sensitive than beamforming demand.
and requires detailed channel information. This is particularly
challenging when channel information related to other cells are The figure illustrates downlink user throughput at very high
needed. Third, the signal processing computations involved in load in a dense city center with (right figure) and without (left
advanced nullforming can be substantial, which is a hardware figure) Interference Sensing enabled. Different colors represent
challenge. different user throughputs, ranging from poor to good. The
network load is increased until the network starts to show signs
of overload using standard reciprocity-based beamforming.
Without Ericsson Interference Sensing enabled, many users with
poor user throughput due to partial overload and high inter-cell
interference are observed. By enabling the Ericsson Interference
Sensing algorithm, it is seen that many user get a much better
user experience due to reduced inter-cell interference.
The gains of Ericsson Interference Sensing will depend on the

use case and deployment, but it is an important addition to the
Massive MIMO toolbox.
Frequency band interworking
The capacity dimensioning examples [TP9.2] showed that The slide illustrates an example with an NR mid-band
adding mid-band gives a substantial downlink capacity and deployment with a supportive NR/LTE low-band deployment.
end-user throughput improvement, but mid-band together In this example, the most limiting physical channel on mid-band
with Massive MIMO is needed to reach superior performance. is uplink data (PUSCH). Given an App coverage requirement of
It was also seen that frequency interworking with a lower band 1 Mbps in the uplink, the link budget of PUSCH is 8 dB worse
is essential when deploying TDD mid-bands from an uplink than the second worst physical channel which is channel state
perspective. This slides illustrates another benefit of frequency information feedback via PUSCH. Channel state information
band interworking, namely the offered coverage extension of feedback via PUSCH has in this example 8 dB worse link budget
mid-band. than downlink synchronization (SSB).
One major benefit of frequency band interworking is the offered By using dual connectivity or carrier aggregation, the limiting
coverage extension of mid-bands. By noticing that different uplink data (PUSCH) can be moved to the low-band with better
physical channels experience different coverage (have different coverage as soon as PUSCH looses coverage on mid-band.
link budgets), weak channels can be moved to a lower band This step increases the overall mid-band link budget by 8 dB.
with better coverage, hence being able to use strong channels in To further increase the mid-band coverage, the cross-carrier
mid-band in a larger area. It is important to notice that the link scheduling functionality offered via carrier aggregation can be
budgets of the different channels depend on many factors, such utilized, hence transmitting also L1 control channel information
as, output powers, bandwidths, receiver sensitivities, antenna on the low-band. Moving channel state information feedback
gains, and required operating points (SINRs), but in many cases related to mid-band PDSCH to low-band gives in this example
downlink channels have better coverage than uplink channels. an additional 8 dB improved mid-band link budget. As SSB is
typically needed on mid-band, the mid-band cannot be extended
by moving/removing SSB. There are, however, other means to
enhance the mid-band link budget, for example, employ power
boosting for SSB.

Trial with Telia Norway gives 64% cell

range increase
Cell edge: Cell edge:
1480 m 2430 m
+ Dual + Carrier
5G (TDD 3.5 GHz) Connectivity Aggregation
4G anchor (FDD 1800 MHz) uplink

data
Our mobile and FWA customers will

5G (FDD 700 MHz)* uplink data experience enhanced capacity, coverage
uplink control and speed… We will also be able to
considerably improve indoor coverage Trial performed in
Lillestrøm, Norway by Telia
*) 5G on FDD 700MHz enabled with ESS (4G part not shown, since not involved in CA call) Espen Weum, Acting Head of Infra, and Ericsson on February
Telia Norway 23rd, 2021
This slide illustrates the coverage extension gains offered by

frequency band interworking in a real trial (5G FDD-TDD Carrier
Aggregation Trial).
In this trial it turns out that carrier aggregation (CA) provides

64% increased cell range, which matches well our link budget
predictions. Note, though, that other assumptions may lead to
other conclusions, e.g. less gain of CA.
5G (TDD 3.5GHz, n78)
4G/5G with Ericsson spectrum sharing (FDD 700MHz, n28)
(4G anchor band 1800 MHz, b3)
More information on Ericsson spectrum sharing (ESS) can

be found in [18].
Codebook-based vs reciprocity-based
beamforming
Codebook-based vs reciprocity-based beamforming No
Switch between codebook and reciprocity

3GPP Urban macro; 3.5 GHz
for superior performance
Capacity gains relative

20
to codebook [%]
10
-10
-20
500 m ISD 1000 m ISD
Reciprocity Reciprocity+codebook
Reciprocity-based Reciprocity-based Codebook-based Codebook-based

MU-MIMO SU-MIMO SU/MU-MIMO SU/MU-MIMO with
FDD-TDD CA
Fundamentally the performance potential of codebook-based Left upper figure illustrates over the air (OTA) results comparing
and reciprocity-based beamforming is similar. What matters is downlink throughput of reciprocity-based and codebook-based
to get sufficient channel state information for the purpose of the beamforming as a function of path loss and shows that there is
transmission (for example, wideband SU-MIMO requires much a switching point where codebook-based performs better than
less channel state information compared to frequency-selective reciprocity-based beamforming.
MU-MIMO). There are, however, many ifs and buts that impact
which beamforming type that works best in practice, see [TP3, p. Right upper figure shows simulation results comparing gains of
38] and [TP7, p. 101-103] for further details. different beamforming algorithms relative to codebook-based
beamforming in 3GPP Urban Macro with 500 m inter-site
This slide illustrates that with current standard and algorithm distance and 1000 m inter-site distance, respectively. It is seen
product choices, reciprocity-based beamforming excels in that reciprocity-based beamforming performs better than
deployments with sufficient good coverage for uplink sounding, codebook-based beamforming for 500 m inter-site distance,
otherwise codebook-based beamforming performs better. whereas for 1000 m inter-site distance, reciprocity-based
This is the reason why Ericsson promotes a solution where shows a loss. Best performance in all deployments and for
the beamforming algorithm is adapted to the situation. The all key performance indicators is given by the combined
beamforming algorithm will essentially be switching between reciprocity+codebook algorithm, which switches between
reciprocity-based MU-MIMO, reciprocity-based SU-MIMO, reciprocity-based and codebook-based beamforming based
codebook-based SU/MU-MIMO, and finally codebook-based on the uplink sounding signal quality. Hence, UEs that have a
SU/MU-MIMO with feedback on a lower frequency band, sounding SNR better than a threshold are using reciprocity-
depending on the coverage situation (illustrated by the lower based beamforming, otherwise codebook-based beamforming.
right figure).
Larger antennas to boost the radio

network performance
Larger antennas to boost the radio network Relativ
Zooma
performance
Frequency 700 2.1 3.7 50
MHz GHz GHz Uplink Downlink 46
40
relative to legacy
Duplex FDD FDD TDD

39
Capacity gain
Bandwidth 20 MHz 20 MHz 100 MHz 30 33

Output power 160 W 160 W 320 W
[%]
20 23 22
10
7
0
3.7 GHz 3.7 GHz
0.7 GHz 2.1 GHz 2.1 GHz
Legacy 32T32R
64TR 64TR
4T4R 384AE 32T32R 384AE
20AE 192AE
192AE
2.1 GHz
4T4R
40AE 3.7 GHz
64TR
192AE
(1) FDD Massive (2) Larger 3.7 (3) FDD Massive MIMO
MIMO antenna and larger 3.7 antenna
As discussed in [TP1, p. 9], the maximum antenna gain increases From a downlink perspective, it is seen that both (1) and (2)
as the size of the antenna grows. The network performance provide similar capacity gains in the order of 20-25%, and
benefits of larger Massive MIMO antennas are currently the gain of combining them (3) is additive, rendering a total
discussed in many contexts, such as: Can uplink performance be of 45% capacity gain. From an uplink perspective, however,
improved by increasing the antenna size? What antenna size is it is clear that (1) provides much larger gains compared to (2)
needed for new frequency bands in between FR1 and FR2? Can and the additional gains from (3) compared to (1) is rather
current radio network performance be kept with reduced output small. Needless to say, this is not a completely fair comparison
power by introducing larger antennas? This could potentially as the antenna improvement relative to legacy in (1) is larger
facilitate a more energy efficient network. See also the than in (2). At the same time, however, 3.7 GHz has much
discussions in [Ch. 2, p. 32] and [Ch. 5.2, p. 96] related to larger more bandwidth than 2.1 GHz, hence a larger performance
antennas and FDD Massive MIMO. potential. For the uplink, the coverage at 3.7 GHz is too poor in
this deployment for a 3 dB coverage improvement to impact
The figure illustrates capacity gains from larger antenna in a the network capacity, and the main problem is to serve the
dense Urban deployment with 500 m inter site distance. Three indoor UEs. As most UEs have uplink coverage on 2.1 GHz,
different frequency bands are employed, 700 MHz, 2.1 GHz and an improvement of this band translates to improved network
3.7 GHz, and the legacy deployment antenna configurations at capacity. For the downlink, coverage is very good for all bands,
these frequencies are 4T4R with 20 antenna elements (AEs), and a twice as large antenna at 3.7 with 100 MHz bandwidth
4T4R with 40 AEs and 64T64R with 192 AEs, respectively. gives in this case similar gains as a five times larger antenna with
Capacity gains relative to legacy for three different antenna Massive MIMO at 2.1 GHz with only 20 MHz bandwidth.
configurations are shown: (1) FDD Massive MIMO at 2.1 Hz
based on 32T32R with 192 AEs. (2) Twice as large antenna at Needless to say, the gain of larger antenna will depend on
3.7 GHz, that is 64T64R with 384 AEs. (3) Both FDD Massive various conditions such as deployment, frequency bands,
MIMO at 2.1 GHz and twice as large antenna at 3.7 GHz. bandwidth, etc., and these results should be seen as indicative.
The main conclusion is to “strive for improving the band that has
the potential to carry most traffic”.
Downlink performance gains of adding

mmWave in a multiband urban macro
deployment
Downlink performance gains of adding mmWave Tw
al
in a multiband urban macro deployment
Baseline mmWave Downlink performance gains by adding mmWave
Frequency band 700 MHz 3.5 GHz 26 GHz (baseline + mmWave)
Bandwidth 10 MHz 100 MHz 400 MHz
• Modest impact on capacity but substantial
throughput improvements for good UEs
Radio/antenna 2T RRU AIR3219 AIR5322
• High mmWave potential for outdoor, whereas
indoor is more challenging
500 457
Gain relative to baseline [%]
450 140%
80% indoor traffic 120%
Capacity gain relative to

400 120%
350 95%
100%
300 254 baseline 80% 62%
250
200 60%
133 35%
150 40%
100 17%
35 20%
50
0%
0
100% 80% 60% 40% 20%
Capacity 5th percentile 50th percentile 95th percentile
UE throughput UE throughput UE throughput Percentage of indoor traffic
As discussed in [Ch. 2, p. 35], adding mmWave has the potential The left figure shows gains in terms of downlink capacity and
to substantially boost the user experience due to the vast amount different downlink UE throughput percentiles. Evidently, the
of spectrum. The main challenge of mmWave is to get good capacity gain is rather modest given the amount of added
enough coverage due to worse propagation properties at higher spectrum; 30% increased capacity by increasing the bandwidth
frequencies. by ~360%. The user experience is, however, substantially
improved for all users and especially for the good users.
This example illustrates downlink network gains achieved by
adding 400 MHz bandwidth at 26 GHz to a baseline network The right figure shows the downlink capacity gain as a function
consisting of 10 MHz at 700 MHz and 100 MHz at 3.5 GHz. of the percentage of indoor traffic. These results show that
The deployment scenario is urban macro with 500 m inter site mmWave has the potential to bring substantial gains for outdoor
distance and 80% indoor traffic as default. A 10 Mbps cell-edge traffic, whereas indoor traffic is more challenging to address with
(5th percentile) UE throughput requirement for the baseline mmWave outdoor macro deployments.
scenario is used to read the capacity and UE percentile gains.
9.4 TCO analysis
TCO considerations for mid-band

deployments
TCO considerations for mid-band deployments
Massive MIMO provides the best total

cost per capacity solution TCO savings per capacity
Old
8T8R antenna,
and radio
mounted
New
behind the
antenna mMIMO
integrated TCO per
radio and capacity
antenna
TCO per
capacity

(8T) solution solution
Total cost of ownership (TCO) calculations depend on many The latest Ericsson Massive MIMO product advancements have
factors that are specific for each service provider and may differ significantly improved the TCO. In fact, the TCO for Massive
from site to site. The following three slides aim to illustrate some MIMO is now comparable with the TCO for a conventional radio
generic key aspects of TCO; see also [Ch. 2, p. 39]. and antenna solution.
Cost can be measured in different ways, but the proposed The capacity benefits of Massive MIMO compared to
metric is TCO per Capacity. Using TCO as a cost measure over conventional antenna solutions depend to a large extent on
the investment cycle is reasonable as it takes both capital the deployment scenario [TP9, p. 143]. However, in many
expenditures (CAPEX) and operational expenses (OPEX) into deployment scenarios, the capacity gain of Massive MIMO over
consideration. Capacity is a measure of the benefits that the a conventional 8T solution is in the order of two to three times.
product brings and is therefore a good measure of value.
Combining the TCO analysis and the predicted capacity benefits,
When making TCO comparisons, all important cost drivers results in that Massive MIMO provides a much better total
should be considered, including radio & antenna equipment, cost per capacity solution in many deployments compared to
RAN compute capacity, site material, rollout & installation, site conventional radio + antenna solutions.
rental and energy consumption. Service providers do usually
rent the sites, and, for example, the site rental may be different The details of the TCO per capacity process is illustrated on the
for the case of adding conventional remote radios with passive following slides.
antennas compared to adding one Massive MIMO product.
Downlink capacity dimensioning

Downlink capacity dimensioning
Massive MIMO facilitates a

future proof superior mid-band
experience
X Traffic density year x
Traffic growth
year-over-year
according to Ericsson
I G J H
mobility report ~50%
A key radio network dimensioning criterion is capacity which Some reflections from the results shown in the figure are as
determines how much traffic the network can sustain with a follows:
maintained user experience (see [TP9, p. 143-144]). Capacity
is defined as traffic per area unit (bits per second per area unit) • Both Massive MIMO variants deliver more than twice the
given a minimum user throughput requirement for the so-called capacity compared to the RRU solutions
cell-edge users (typically 5th user percentile). Here a throughput
• It is seen that all radio solutions provide good downlink
requirement of 50 Mbps is assumed to reach a superior 5G
throughputs well above the stipulated downlink throughput
experience.
requirement of 50 Mbps in today’s network (represented by
The figure illustrates the downlink user throughput for cell-edge the point 0), as well as after 3 years (point 3)
users as a function of the traffic density for different radio • After 5 years expansion (point 5), there is, however, a dramatic
products in an urban mid-rise deployment scenario. The four difference between RRU solutions and Massive MIMO
different radio products are RRU 4T4R, RRU 8T8R, Massive solutions. In particular, the 4T4R solution will no longer fulfill
MIMO 32T32R and Massive MIMO 64T64R. the throughput requirement and the network needs to be
upgraded. The 8T8R solution is still meeting the requirement
A network capacity dimensioning analysis needs to consider the
but delivers significantly reduced quality compared to the two
traffic increase over time. Given today’s assumed traffic density
represented by the point 0, the different points 3, 5 and 7 in the
figure represent the predicted capacity demand after 3, 5, and 7, • The Massive MIMO solutions, at least the 64T64R, can sustain
years, respectively. the capacity demand for 7 years before a network capacity
upgrade is needed.
From this capacity dimensioning example, it is clear that the

capability of different radio solutions to handle the expected
capacity demand of the future can be very different. In this
example Massive MIMO can handle the capacity requirements
up to 7 years while the RRU radio solutions can handle at most
4-5 years.
Structure of TCO analysis

Structure of TCO analysis U
A
TCO for infrastructure OPEX CAPEX

Cost per site TCO per site
Small differences in TCO for remote
radios and Massive MIMO

Capacity gain
Capacity gain
Capacity needs vs. product potential
Massive MIMO efficiency is scenario
dependent
TCO/GB
TCO/GB
Cost per delivered capacity
Massive MIMO has superior total cost
per capacity
A per-site total cost of ownership (TCO) analysis is an important Third step is then to normalize the TCO results from the first step
step in the process of deciding a recommended radio solution. with the predicted capacity gain from the second step. As the
TCO calculations depend on many local and CSP specific TCO for Massive MIMO is only slightly higher than for RRU, but
factors. This slide illustrates three steps of the TCO analysis that Massive MIMO provides roughly twice the capacity compared to
highlights that what matters most for investment decisions is the RRU, the “TCO per capacity” is reduced with more than 40%
TCO per delivered capacity. if Massive MIMO is used.
First step is to run a traditional TCO analysis of the cost parts, Section [TP9.2] illustrated that adding mid-band with Massive
where both capital expenditures (CAPEX) and operational MIMO provides a future-proof deployment with superior
expenses (OPEX) are evaluated over a typical depreciation time. performance, and here it is further illustrated that Massive MIMO
Although the upfront CAPEX for Massive MIMO is higher than often has superior total cost per capacity compared to other
for conventional RRU solutions, this step often demonstrates solutions. Hence, in many cases Massive MIMO is a cost-efficient
that the TCO difference between deploying mid-band with a solution for mid-band deployments in many cases.
conventional RRU and a more advanced Massive MIMO solution
is rather small. This is mainly because the conventional solution
has higher cost related to among others network rollout, site
rental and other parts.
Second step is to determine the offered capacity of the products.

As illustrated in [Ch. 2, p. 39], the capacity benefits of Massive
MIMO over a conventional solution depend on many aspects.
Still, in many deployment scenarios, see previous slide, it is
reasonable to assume ~2-3 times more capacity from Massive
MIMO than a conventional system. The figure illustrates that 64T
Massive MIMO provides twice the capacity compared to 8T RRU
in an urban mid-rise deployment scenario.
Summary of radio network performance

Summary of radio network performance No chan
Wrong fo
• Deployment environment, array structure
and algorithm choices are key components
affecting the performance of Massive MIMO
• Deploying mid-band with Massive MIMO

enables superior performance in a cost-
efficient way
• Adding mmWave will further boost

performance by offloading lower bands
and improving end-user experience for users
in good radio conditions
• Good match between our simulation results
and real network performance
Radio network performance depends on many aspects that Deep-dives into a few selected areas that are subject to
interact in an intricate manner, and it is important to understand frequently asked questions were given. Effects of the number of
the driving mechanisms and have models that capture the MU-MIMO layers, gains from Interference Sensing, benefits of
reality sufficiently well while still being computationally frequency band interworking, and the impact of type of channel
efficient. Ericsson is in the forefront in the area of radio network state acquisition approach (codebook-based vs. reciprocity-
performance evaluations using state-of-the-art knowledge and based) were discussed. Furthermore, the benefits of larger
models. The predictions based on simulations have proven to antennas and performance gains from mmWave deployments
match real network performance on countless occasions. were illustrated.
Several illustrations of that Massive MIMO deployments Performance evaluations can guide service providers when it
facilitate superior radio network performance have been given. comes to network evolution and product needs. They are also
In particular, key aspects impacting Massive MIMO performance, vital in the R&D flow. It provides an understanding of how
such as antenna configuration, deployment scenario and different deployments, configurations and algorithms perform
beamforming algorithm choice, have been discussed and before making costly R&D product investments.
illustrated. A detailed illustration of capacity dimensioning
benefits of Massive MIMO over other radio solutions has been
provided.
170 Ericsson | The Massive MIMO Handbook
10. Abbreviations
10. Abbreviations
Ericsson | The Massive MIMO Handbook 10. Abbreviations 171
3GPP 3rd generation partnership project EPO energy performance optimizer
2D two dimensional FCC federal communications commission
AAS advanced antenna systems; note: FDD frequency division duplex

synonymous to Massive MIMO
FH frequency hopping
ACK acknowledgement (positive)
GHz giga Hertz
AI artificial intelligence
GP guard period
AIR antenna integrated radio
gNB node B (NR)
AOSA array of subarrays
HARQ hybrid automatic repeat request
AR augmented reality
HW hardware
AS antenna switching
Hz hertz
BF beamforming
IBW instantaneous bandwidth
BB baseband
IEEE institute for electrical and electronics
BW bandwidth engineers
CA carrier aggregation IRC interference rejection combining
CAPEX capital expenditures ISD inter-site distance
CC component carrier MAC CE medium access control channel element
CCH control channel MBB mobile broadband
CO2 carbondioxide Mbps mega bit per second
CoMP coordinated multi-point MIMO multiple input multiple output
CQI channel quality indicator ML machine learning
CSP communication service provider mmWave milimeter wave
CSI channel state information MCG master cell group
CSI-RS CSI-reference signal MSGx message x
DC dual connectivity MU-MIMO multi-user MIMO
DCI downlink control information NACK negative ACK
DFT discrete fourier transform NR new radio
DL downlink OPEX operational expenditures
DMRS demodulation reference signal ORAN open RAN architecture
EIRP equivalent isotropic radiated power OSI open systems interconnection
EIS equivalent isotropic sensitivity PCell primary cell
EM electro magnetic PSCell primary SCG cell
EMF electro magnetic field PBCH physical broadcast channel
10. Abbreviations
172 Ericsson | The Massive MIMO Handbook
PDCCH physical downlink control channel SIB x system information block x
PDSCH physical downlink shared channel SINR signal-to-noise-and-interference ratio
PMI precoding matrix indicator SNR signal-to-noise ratio
PRACH physical random access channel SON self-organizing network
PUCCH physical uplink control channel SUL supplementary uplink
PUSCH physical uplink shared channel SpCell special cell
PSK phase shift keying SR scheduling request
PSS primary synchronization signals SS synchronization signals
QAM quadrature amplitude modulation SRS sounding reference signal
QPSK quadrature PSK SSB block containing SS and PBCH
QoS quality-of-service SSS secondary synchronization signal
Qx quarter x; x=1,2,3,4 SU-MIMO single-user MIMO
R receiver radio chain SW software
RAN radio access network T transmitter radio chain
RAN1 3GPP working group for radio layer 1 TCBW total configured bandwidth
(physical layer)
TCO total cost of ownership
RAN2 3GPP working group for radio layer 2
and radio layer 3 radio resource control TDD time division duplex
RAN4 3GPP working group for UTRAN/E- TRP transmission point

UTRAN/NG-RAN architecture and
related network interfaces UCI uplink control information
RAN4 3GPP working group for radio UE user equipment

performance and protocol aspects
UL uplink
RAN5 3GPP working group for mobile terminal
conformance testing VR virtual reality
RAR random access response XR extended reality
RMSI remaining system information (SIB1) YoY year-on-year
RI rank indicator
RRC radio resource control
RF radio frequency
SCell secondary cell
SCG secondary cell group
SCS subcarrier spacing
10. Abbreviations
Ericsson | The Massive MIMO Handbook — Technology Primer Chapter title 173
Ericsson.com The content of this document is subject EN/LZT 4/28701-FGB1010987 Uen Rev C
to revision without notice due to © Ericsson AB 2024
continued progress in methodology,
design and manufacturing. Ericsson
shall have no liability for any error or
damage of any kind resulting from the
use of this document

Massive Mimo Handbook 3rd Edition

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Massive Mimo Handbook 3rd Edition

Uploaded by

Copyright:

Available Formats

third edition

This book contains two documents:

• Massive MIMO Handbook

EN/LZT 4/28701-FGB1010987 Uen Rev C

Differences between ​the 1st edition and the 2nd edition

Massive MIMO explores the spatial domain by using

Dense urban high-rise

Capacity Compact Coverage

Executive Summary—finding suitable

Massive MIMO for capacity

Capacity Remote radio due to higher

Outline and reading guidelines

1. Introduction to Massive MIMO 10

2. Requirements and solutions

3. Massive MIMO solutions 42

4. Guiding principles for how to select

1. Introduction to Massive MIMO

• What Massive MIMO is

• Why Massive MIMO is used

• How Massive MIMO works

1. Introduction to Massive MIMO

Massive MIMO is here and now E

40% 88% 83%

Note that in this example, the deployment is made in an

1. Introduction to Massive MIMO

Massive MIMO solution

Massive MIMO solution

Baseband Massive MIMO radio Massive MIMO

Baseband Remote radio Passive Conventional

1. Introduction to Massive MIMO

Massive MIMO is addressing key

Total cost of ownership

User Coverage Distance

1. Introduction to Massive MIMO

Massive MIMO offers a diverse set

Beamforming MIMO – Spatial multiplexing

These fundamental technology components can be combined

1. Introduction to Massive MIMO

The fundamental massive MIMO techniques

Single-User MIMO Multi-User MIMO

Further performance boosters

Main performance tools​ ​Further performance boosters​

1. Introduction to Massive MIMO

A Massive MIMO radio uses antenna arrays

Small array Conventional system Massive MIMO

1. Introduction to Massive MIMO

Massive MIMO — future-proof

Mbps Downlink user throughput illustration

250 FDD bands FDD + NR TDD Conventional

Downlink user throughput

FDD + NR TDD Massive MIMO

1. Introduction to Massive MIMO

Summary — Massive MIMO improves

● Massive MIMO has reached general use

1. Introduction to Massive MIMO

2. Requirements and solutions in an evolving 5G network

• Network variations – each site is unique

● Spectrum properties affecting the radio

2. Requirements and solutions in an evolving 5G network

A new 5G experience No chan

Example 1: Extended reality services Example 2: Industry automation services

2. Requirements and solutions in an evolving 5G network

Higher capacity is required in networks

Differences between the 1st edition and the 2nd edition

Main performance tools Further performance boosters