Massive MIMO for 5G below 6 GHz

Achieving Spectral Efficiency, Link Reliability, and

Low-Power Operation

Associate Professor Emil Björnson

Department of Electrical Engineering (ISY)

Linköping University
Sweden
 Dr. Emil Björnson
• PhD from KTH, Sweden; Postdoc at SUPELEC, Paris, France
• Associate professor at Linköping University, Sweden
• 10 year experience of MIMO research
• 2 books and 7 best paper awards
• Some ten pending patent applications Foundations and Trends® in

FnT SIG 11:3-4

Signal Processing
Massive MIMO Networks

Writer at the Massive MIMO blog, http://massive-mimo.net

11:3-4

• Spectral, Energy, and Hardware Efficiency

Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency

Emil Björnson, Jakob Hoydis and Luca Sanguinetti
Massive multiple-input multiple-output (Massive MIMO) is the latest technology that will

Massive MIMO Networks

First author of textbook “Massive MIMO Networks”, Nov. 2017
improve the speed and throughput of wireless communication systems for years to come.
Whilst there may be some debate over the origins of the term Massive MIMO and what it

• precisely means, this monograph describes how research conducted in the past decades lead
to a scalable multiantenna technology that offers great throughput and energy efficiency under Spectral, Energy,
practical conditions.

Written for students, practicing engineers and researchers who want to learn the conceptual
and Hardware Efficiency
and analytical foundations of Massive MIMO as well as channel estimation and practical
considerations, it provides a clear and tutorial-like exposition of all the major topics. The Emil Björnson, Jakob Hoydis

Outline
monograph contains many numerical examples, which can be reproduced using Matlab code
that is available online at https://dx.doi.org/10.1561/2000000093_supp. and Luca Sanguinetti
Massive MIMO Networks is the first monograph on the subject to cover the spatial channel
correlation and consider rigorous signal processing design essential for the complete
understanding by its target audience.

“Massive MIMO is an essential topic in the field of future cellular networks. I have not seen any
other [work] which can compete at that level of detail and scientific rigor. [It] will be very useful

1. Why Cellular Networks Must Become More Efficient

to PhD students and others starting in this area. [The] reading [is] particularly pleasant and rich.
Overall, a great tool to researchers and practitioners in the field.” — David Gesbert, EURECOM

Emil Björnson et al.

“The authors provide an enlightening introduction to the topic, suitable for graduate students
and professors alike. Of particular interest, the [monograph] provides an updated assessment
of the performance limiting factors, showing for example that pilot contamination is not a

2. How Massive MIMO Improves Spectral Efficiency

fundamental limitation.” – Robert W. Heath Jr., The University of Texas at Austin

This book is originally published as

Foundations and Trends® in Signal Processing
Volume 11 Issue 3-4, ISSN: 1932-8346.

3. Beyond Mobile Broadband: now

now
the essence of knowledge

2 Link Reliability, Low-Power Operation

 The Success of Wireless Communications

• More devices and data traffic every year How to pay

• 10% more devices for this?
• 47% more traffic (33% more per device) Higher network
throughput in 5G
Revenue from
Connected devices Data traffic new use cases:
Billion Exabyte/month 9.1 GB/month/person
3.6/person
80 Internet-of-things
30
2/person 60 Ultra-reliable
20 970 MB/month/person
communication
40
10 20 etc.

0 0
2016
2017
2018
2019
2020
2021
2022

2016
2017
2018
2019
2020
2021
2022
Data source:
Ericsson Mobility Report
3 (July, November 2017)
 Improving Cellular Networks
• Formula for Network Throughput [bit/s/km2]:
Throughput = Cell density 8 Available spectrum 8 Spectral efficiency
bit/s/km/ Cell/km/ Hz bit/s/Hz/Cell

• Two-Tier Networks
• Hotspot tier
• High cell density, short range per cell
• Wide bandwidths in mm-wave bands
• Spectral efficiency less important
• Coverage tier (focus today)
• Provide coverage, elevated base stations 3.4-3.8 GHz primary
• Outdoor-to-indoor coverage: Operate <6 GHz 5G band in Europe
and elsewhere
4 • High spectral efficiency is desired
 Interference Limits the Spectral Efficiency
Base stations

8 8

Spectral Efficiency [bit/s/Hz]

Spectral Efficiency [bit/s/Hz]

6 6

4 4

2 2

0 1000 0 1000
1000 1000
500 500
500 500
• dffd
Position [m] 0 0 Position [m] Position [m] 0 0 Position [m]

Cell densification is not a solution

Mediocre performance
Pathloss exp: 3
at most places!
5
Higher frequencies makes it worse Cell edge: 5-10 dB
 How to Achieve More Uniform Coverage?

8 8

Spectral Efficiency [bit/s/Hz]

Spectral Efficiency [bit/s/Hz]

6 6

4 4

2 2

0 1000 0 1000
1000 1000
500 500
500 500

• dffd
Position [m] 0 0 Position [m] Position [m] 0 0 Position [m]

Desired: Stronger signal, same interference levels

6
 Beamforming is the Solution!

User User User

Signal goes in
all directions
Substantial
side-lobes Main
Tiny Narrow
lobe
side-lobes main lobe

More antennas

Same transmit power More antennas

• Color indicates path loss in dB • Narrower beams, laser-like

• M base station (BS) antennas • Array gain: 10 log10(M) dB larger at user

• Main lobe focused at user • Less leakage in undesired directions

7
 Massive MIMO (multiple input multiple output)

• Main Characteristics
• Many BS antennas; e.g., M = 200 antennas, K = 40 single-antenna users
• Many more antennas than users: M≫K
• High spectral efficiency
• Many simultaneous users
• Strong directive signals
• Little interference leakage

Seminal work: Thomas L. Marzetta, “Noncooperative Cellular Wireless with Unlimited

Numbers of Base Station Antennas,” IEEE Trans. Wireless Communications, 2010
• Combines the best concepts from past decades of multi-user MIMO research
8
• 2013 IEEE Marconi Prize Paper Award, 2015 IEEE W. R. G. Baker Award
 Massive MIMO Provides Favorable Propagation

• Consider two users

𝐡B 𝐡D
• M-dimensional channels: 𝐡B , 𝐡D
Favorable propagation
𝐡G 𝐡/
and are orthogonal Base station can fully separate the users
𝐡G 𝐡/

|𝐡BF 𝐡D |
𝐡B 𝐡D

Source: J. Hoydis, C. Hoek, T. Wild, and S. ten Brink,

9 “Channel Measurements for Large Antenna Arrays,” ISWCS 2012
 Deploying Many Antennas Below 6 GHz
One dual-polarized
antenna elements
Look
inside Number of Antennas
• 8·8 = 64 per sector
• 192 antennas per site

LTE: One input/output per polarization!

3 sectors, Massive MIMO: One per antenna element
8-antenna LTE-A
Upgrade Existing Sites to Massive MIMO
1 site No sectorization (achieved by beamforming)
Equipment size similar to top-of-the-line LTE
Massive in numbers, not in size
One dual-polarized
10
antenna panel
 Spatial Multiplexing Requires Digital Beamforming

• How to implement beamforming?

• Send same signal from all antennas
• Vary phase/amplitude per antenna
• Vary phase/amplitude per subcarriers

• Spatial multiplexing: Superimpose beams

Flexible Implementation
Hybrid beamforming:
Cannot adapt amplitude or subcarriers
Digital beamforming: Full flexibility
Digital is the future!

11
Digital beamforming Hybrid beamforming
“Channel State Information isn’t Everything; it’s the Only Thing” – T. Marzetta
We need to know where the point the beam!

• Conventional approach: Grid-of-beams

• Try 8 angular beams, user reports the best one
• Good: Simple, works with both TDD and FDD
• Bad: Never a perfect match; too much inter-user interference

• Massive MIMO: Uplink estimation

• User sends pilot signal, BS estimates channel
• Good: Well-matched estimates, scalable with many antennas
• Bad: Only works in TDD, where uplink estimates useful for downlink

12
FDD = Frequency-division duplex, TDD = Time-division duplex
 World Record in Spectral Efficiency

• 145.6 bit/s/Hz/cell
• Set jointly by researchers
in Bristol and Lund, 2016
• 128 BS antennas
• 22 single-antenna users
• 256-QAM signals
• 20 MHz band at 3.5 GHz

Is this practical?
Screenshot from “Massive MIMO World Records”
Multiplexing tens of users is practical
Link: https://youtu.be/NoDP3g8XHVQ
Low-order modulations will mainly be
used in practice
13
 High Spectral Efficiency in Cellular Networks

Uplink simulation: SNR 5 dB, i.i.d. Rayleigh fading,

zero-forcing combining, channels fixed for 500 channel uses
Pilots reused in
every third cell High spectral efficiency per cell, ~3 bit/s/Hz to every user
14
 5G is More Than Broadband: Internet-of-things (IoT)

• Wirelessly connected society

• Machines, vehicles – everything gets connected
• Other use cases than mobile broadband

• Case 1: Link Reliability is Very Important

• Connected factory robots, traffic safety applications, etc.
• Ultra-reliable low-latency communication (URLLC)

• Case 2: Massive machine-type communication (mMTC)

• Many low-cost sensors and actuators deployed everywhere (50 billion by 2020)
• Sporadic transmission, battery should last for 5 years

Can Massive MIMO play a role here?

15
 Channel Hardening
Consider a random channel, e.g., 𝐡 ∼ 𝐶𝑁(𝟎, 𝐈N )
𝐡
Variations of effective channel reduce with 𝑀:
1 Mean: 1
𝐡 D has R Few antennas
𝑀 Variance: 1/𝑀

3 Narrower beam:
Mean value Fewer multipath
2.5
Percentiles components involved
One realization
2
Many antennas
90%
1.5 D D
𝐡 ≈E 𝐡
1
10% Double benefits:
0.5
D
𝐡 scales with 𝑀
0
0 50 100 150 200 250 300 350 400 Variations reduces
Number of BS Antennas (M)
16
 Great Link Reliability and Simplified Resource Allocation

Higher reliability, lower latency Resource allocation made simple

10 3
M=100 Conventional Massive MIMO

Frequency

Frequency
10 2
Channel gain

10 1 M=1

10 0

10 -1

10 -2
0 20 40 60 80 100 Time Time
Realization

D • All subcarriers good, all the time

• Lost package if 𝐡 < threshold
• No need to schedule based on fading
• Less likely with channel hardening
• Each user gets the whole bandwidth,
• Fewer retransmissions whenever needed
17
 Two Ways to Exploit the Array Gain

1) Range Extension 2) Low-Power Operation

Use 5 log10(M) to 10 log10(M) dB less power

10 log10(M) dB

• Same range with reduced power

• Use same transmit power
• Increase battery lifetime in uplink
• Higher rates to already covered places
• Low power per antenna in downlink
• Reach new places (e.g., indoor) 40 W à 4 W per BS, 40 mW/antenna
18
 Supporting Internet-of-Things (IoT)

Base station Sensor

• SNR over 100 kHz channel:

20 dBm + 2.15 dBi + 2.15 dBi − 150 dB − −120 dBm = −5.7 dB
Transmit power Antenna gains Pathloss Noise power

• Sufficient for binary modulation with repetition coding

• Transmit a few data packages per day (very low energy per package)

Massive MIMO with M = 100

Increase SNR by 20 dB (range extension) Reduce transmit power to 10 dB
Improve link reliability Up to 10x longer battery life
19
 Summary: Massive MIMO for 5G below 6 GHz

1. Mobile broadband applications

• Very high spectral efficiency, multiplex many users
• Great improvements at the cell edge

2. Ultra-reliable low-latency communication (URLLC)

• Channel hardening alleviates small-scale fading
• Fewer retransmissions, more predictable performance

3. Massive machine-type communication (mMTC)

• Extend coverage, more cost-efficient deployment
• Reduce transmit power for battery-power devices
20
 Learn More: Blog and Book

• Massive MIMO blog: www.massive-mimo.net

New book: Foundations and Trends® in

FnT SIG 11:3-4

Signal Processing

Youtube channel: Massive MIMO Networks

Spectral, Energy, and Hardware Efficiency
11:3-4

Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency

Emil Björnson, Jakob Hoydis and Luca Sanguinetti

Emil Björnson, Jakob Hoydis

Massive multiple-input multiple-output (Massive MIMO) is the latest technology that will
improve the speed and throughput of wireless communication systems for years to come.
Massive MIMO Networks
and Luca Sanguinetti (2017),
Whilst there may be some debate over the origins of the term Massive MIMO and what it
precisely means, this monograph describes how research conducted in the past decades lead
to a scalable multiantenna technology that offers great throughput and energy efficiency under
practical conditions.
Spectral, Energy,
and Hardware Efficiency
“Massive MIMO Networks:
Written for students, practicing engineers and researchers who want to learn the conceptual
and analytical foundations of Massive MIMO as well as channel estimation and practical
considerations, it provides a clear and tutorial-like exposition of all the major topics. The Emil Björnson, Jakob Hoydis
monograph contains many numerical examples, which can be reproduced using Matlab code

Spectral, Energy, and

that is available online at https://dx.doi.org/10.1561/2000000093_supp. and Luca Sanguinetti
Massive MIMO Networks is the first monograph on the subject to cover the spatial channel
correlation and consider rigorous signal processing design essential for the complete

Hardware Efficiency”
understanding by its target audience.

“Massive MIMO is an essential topic in the field of future cellular networks. I have not seen any
other [work] which can compete at that level of detail and scientific rigor. [It] will be very useful
to PhD students and others starting in this area. [The] reading [is] particularly pleasant and rich.
Overall, a great tool to researchers and practitioners in the field.” — David Gesbert, EURECOM

Emil Björnson et al.

“The authors provide an enlightening introduction to the topic, suitable for graduate students

517 pages, Matlab code

and professors alike. Of particular interest, the [monograph] provides an updated assessment
of the performance limiting factors, showing for example that pilot contamination is not a
fundamental limitation.” – Robert W. Heath Jr., The University of Texas at Austin

This book is originally published as

Foundations and Trends® in Signal Processing
Volume 11 Issue 3-4, ISSN: 1932-8346.

now

now
$40 for paperback until Jan 31 the essence of knowledge

Use discount code 996889

https://youtu.be/m9wEAucKoWo on nowpublishers.com massivemimobook.com
21
 Special offer

Thank you!
Emil Björnson, Jakob Hoydis
and Luca Sanguinetti (2017),
“Massive MIMO Networks:
Spectral, Energy, and
Hardware Efficiency”
Questions are most welcome! Foundations and Trends® in

FnT SIG 11:3-4

Signal Processing
Massive MIMO Networks 11:3-4
Spectral, Energy, and Hardware Efficiency

Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency

Emil Björnson, Jakob Hoydis and Luca Sanguinetti

Massive multiple-input multiple-output (Massive MIMO) is the latest technology that will
improve the speed and throughput of wireless communication systems for years to come.
Whilst there may be some debate over the origins of the term Massive MIMO and what it Massive MIMO Networks
precisely means, this monograph describes how research conducted in the past decades lead
to a scalable multiantenna technology that offers great throughput and energy efficiency under
practical conditions.
Spectral, Energy,
Written for students, practicing engineers and researchers who want to learn the conceptual
and Hardware Efficiency
and analytical foundations of Massive MIMO as well as channel estimation and practical
considerations, it provides a clear and tutorial-like exposition of all the major topics. The Emil Björnson, Jakob Hoydis
monograph contains many numerical examples, which can be reproduced using Matlab code
that is available online at https://dx.doi.org/10.1561/2000000093_supp. and Luca Sanguinetti

Dr. Emil Björnson

Massive MIMO Networks is the first monograph on the subject to cover the spatial channel
correlation and consider rigorous signal processing design essential for the complete
understanding by its target audience.

“Massive MIMO is an essential topic in the field of future cellular networks. I have not seen any
other [work] which can compete at that level of detail and scientific rigor. [It] will be very useful
to PhD students and others starting in this area. [The] reading [is] particularly pleasant and rich.
Overall, a great tool to researchers and practitioners in the field.” — David Gesbert, EURECOM

Emil Björnson et al.

“The authors provide an enlightening introduction to the topic, suitable for graduate students
and professors alike. Of particular interest, the [monograph] provides an updated assessment
of the performance limiting factors, showing for example that pilot contamination is not a
fundamental limitation.” – Robert W. Heath Jr., The University of Texas at Austin

This book is originally published as

Foundations and Trends® in Signal Processing
Volume 11 Issue 3-4, ISSN: 1932-8346.

now

now
the essence of knowledge

Slides, papers, and code available online: $40 for paperback version
Price available until Jan. 31
http://www.ebjornson.com/research
Discount code 996889
http://www.massivemimobook.com Only at nowpublishers.com
 BACKUP SLIDES

23
 Classical Multi-User MIMO vs. Massive MIMO

Classic multi-user MIMO

Classic multi-user MIMO Massive MIMO (Canonical)
Antennas
Antennas users
𝑀,𝑀, users
𝐾𝐾 𝑀𝑀≈≈𝐾𝐾 𝑀≫𝐾
Signal
Signal
processing
processing Non-linear
Non-linearis preferred
is preferred Linear is near optimal
Duplexing
Duplexing
mode
mode Designed for TDD and FDD
Designed for TDD and FDD Designed for TDD w.
reciprocity
Instantaneous channel
Instantaneous channel Known at BS and user
Known at BS and user Only needed at BS
(hardening)
Channel quality
Channel quality Affected by frequency-
Affected by frequency- Almost no channel quality
selective
selective
and fast fading
and fast fading variations (hardening)
Variations in user load
Variations in user load Scheduling needed if 𝐾
Scheduling needed if 𝐾>>𝑀 Scheduling seldom needed
𝑀
Resource
Resource
allocation
allocation Rapid
Rapiddue to fading
due to fading Only on a slow time scale
Cell-edge
Cell-edge
performance
performance Only good if BSs cooperate
Only good if BSs cooperate
FDD = Frequency-division duplex, Improved by array gain of 𝑀
TDD = Time-division duplex

BS cooperation
BS cooperation Highly beneficial
Highly beneficial
if rapid
if rapid Only long-term coordination
24
 MAMMOET (Massive MIMO for Efficient Transmission)
2014-2016

• Bridged many gaps between theoretical and practice

• Testbed demonstrations (real-time operation, mobility)
• New channel models
• Concepts for efficient analog/digital hardware implementation
• Deliverables available: https://mammoet-project.eu

Partners:

25
 Pilot Contamination has Been Blown Out of Proportions
Pilots reused across cells
Interference contaminates estimates
Makes channels unfavorable

b) No limit, asymptotically contamination-free

2012: Caire et al.
Spectral efficiency [bit/s/Hz]

2013: Gesbert et al.

Special case:
c) No limit, but One-ring model
contamination has effect

2017: Björnson et al.

Any nontrivial channel
with spatial correlation
a) Upper limit
2010: Marzetta
0 Special case:
10
1
10
2 3
10 i.i.d. Rayleigh fading
26
Number of antennas (M)
 More important things than
Open Problems “pilot contamination”!
• Make Massive MIMO work in FDD mode
• Long-standing challenge. Is it practically feasible to exploit sparsity?

• Channel measurements, channel modeling, traffic modeling

• Required for system level simulations

• Implementation-aware algorithmic design

• Implement ZF with MR-like complexity. Utilize low-resolution hardware.

• Cross-layer design
• Scalable protocols for random access, control signaling, scheduling

• New deployment characteristics

27 • Multi-antenna users, distributed arrays, cell-free (network MIMO)