TSKS14
Multiple Antenna
Communications
Lecture 1, 2020
Emil Björnson
TSKS14 Multiple Antenna Communications 2020-03-27 2
Three main benefits of multiple antenna communications
1. Beamforming gain
2. Spatial diversity
3. Spatial multiplexing
The course will cover all of them:
• First historical perspective
• Then the theory behind them
TSKS14 Multiple Antenna Communications 2020-03-27 3
Constructive and destructive interference
• One signal:
2 sin 2𝜋𝑡
4 5
(Normalization: ∫3 2 sin 2𝜋𝑡 𝑑𝑡 = 1)
• Two signals:
sin 2𝜋𝑡 + sin(2𝜋𝑡 − 𝜃)
Power depends on the phase difference 𝜃
(time delay)
TSKS14 Multiple Antenna Communications 2020-03-27 4
Omni-directional radiation
𝑀 = 1 antenna
Powers spread out as a sphere
• Received power decays as
1/distance2
TSKS14 Multiple Antenna Communications 2020-03-27 5
Directive radiation – Beamforming gain
𝑀 = 6 antennas
Power directed towards receiver
• Constructive interference Beamforming
• Send signals with time delays
• Needs to know “direction”
Two options:
• 6 times larger received power
• Use 1/6 of power to achieve
same received power
TSKS14 Multiple Antenna Communications 2020-03-27 6
History: Beamforming
• Directional communication
• E.F.W. Alexanderson, ”Transatlantic Radio Communication”,
Trans. AIEE, 1919.
• Directive radiation pattern
• Constructive/destructive interference patterns
• Energy savings or better performance
Called beamforming gain
TSKS14 Multiple Antenna Communications 2020-03-27 7
Multipath Propagation
sin(2𝜋𝑡 − 𝜃)
Transmitter sin 2𝜋𝑡 Receiver
Multipath fading:
Received signal: sin 2𝜋𝑡 + sin(2𝜋𝑡 − 𝜃) Add construtively or
destructively depending on 𝜃
TSKS14 Multiple Antenna Communications 2020-03-27 8
History: Spatial diversity
• Diversity combining
• D.G. Brennan, “Linear diversity combining
techniques,” Proc. IRE, 1959
• Protect against fading/noise
• Multiple antennas with independent
observations
• Small risk that multiple observations are all bad
• Improve reliability
Called diversity gain
TSKS14 Multiple Antenna Communications 2020-03-27 9
Cellular networks
• Designed for mobile telephone systems
• Bullington, K. (1953). “Frequency economy
in mobile radio bands”. The Bell System
Technical Journal.
• Schulte, H. J. and W. A. Cornell (1960).
“Multi-area mobile telephone system”. IEEE
Trans. Veh. Technol.
• Reuse of spectrum in space:
• Densify as usage increases
• Control interference by fractional
spectrum reuse, power control
TSKS14 Multiple Antenna Communications 2020-03-27 10
History: Space division multiple access (SDMA)
• Multiple user communication
• S. C. Swales et al., “The Performance Enhancement of
Multibeam Adaptive Base-Station Antennas for Cellular
Land Mobile Radio Systems” Trans. on Vehicular
Technology, 1990.
• Spatial multiplexing of users
• Serve multiple users on same time and frequency
• Handle more users per base station
• Exists in 4G/5G and Wi-Fi
Called multiplexing gain
TSKS14 Multiple Antenna Communications 2020-03-27 11
History: Point-to-point multi-antenna links
• Point-to-point multiple-input
multiple-output (MIMO)
• G. G. Raleigh and J. M. Cioffi, “Spatio-
temporal coding for wireless
communications,” Globecom 1996.
• Multiple streams per user
• Increase capacity (bit/s) per user
• Requires multiple paths
• Exists in 4G/5G and Wi-Fi
Also called multiplexing gain
TSKS14 Multiple Antenna Communications 2020-03-27 12
Outline of the course
• Lecture 1:
• History of multiple antenna communications
• Channel capacity for single-antenna channels Based on lecture note document:
“Introduction to Multiple Antenna
• Basic multi–antenna channels (Lecture 2) Communications”
• Diversity and ergodic capacity (Lecture 3)
• Point-to-point MIMO (Lecture 4)
• Multi-user MIMO (Lectures 5-12) Based on the book
• Capacity bounds, Channel estimation, “Fundamentals of Massive MIMO”
Power control, etc.
TSKS14 Multiple Antenna Communications 2020-03-27 13
Example: Additive White Gaussian Noise (AWGN) Channel
𝑥 + 𝑦 =𝑥+𝑛
• Noise: 𝑛 ∼ 𝑁(0, 𝑁3/2)
• Energy per symbol: 𝐸 𝑥 5 ≤𝑞
• 𝑥, 𝑦, 𝑛 real valued
TSKS14 Multiple Antenna Communications 2020-03-27 14
How to measure performance?
• Data packet:
Information symbols
• Characterized by
• How many symbols the packet contains
• How many information bits these symbols represent
(determined by the modulation and coding scheme)
• Probability of incorrect decoding at the receiver
TSKS14 Multiple Antenna Communications 2020-03-27 15
Channel capacity
• Channel capacity
• Random variables 𝑋 and 𝑌 𝑥 Channel 𝑦
• Channel described by conditional distribution
𝑝P|R 𝑦|𝑥
• Channel coding theorem
• 𝐶 [bit/channel use] is the capacity of the channel if:
For any given 𝛿 > 0 and 𝛾 > 0,
there exist a channel coding codebook of a finite length 𝑁 that has rate
𝑅 = 𝐶 − 𝛿 and offers an error probability 𝑃 error ≤ 𝛾
TSKS14 Multiple Antenna Communications 2020-03-27 16
Passband and baseband signals
• Bandwidth: Distance from smallest to largest frequency
𝑓 𝑓
𝐵 𝑓_ 𝐵 𝐵 0 𝐵
𝑓_ − 𝑓_ + − +
2 2 2 2
^
Bandwidth (passband): 𝐵 Bandwidth (baseband):
5
Real-valued signal Complex-valued signal
Described by 2𝐵 real samples/s Described by 𝐵 complex samples/s
These samples are used to convey information
TSKS14 Multiple Antenna Communications 2020-03-27 17
Gaussian Distribution
• 𝑥 is zero-mean Gaussian distributed, 𝑥 ∼ 𝑁(0, 𝜎 5)
• Probability density function (PDF): 0.4
1 cd 0.35
b d
𝑝R 𝑥 = 𝑒 5e
2𝜋𝜎 5 0.3
0.25
0.2
0.15
0.1
Properties: 𝐸 𝑥 = 0, 𝐸 𝑥 5 = 𝜎 5 0.05
0
-5 0 5
TSKS14 Multiple Antenna Communications 2020-03-27 18
Complex Gaussian Distribution
• Consider independent 𝑥f , 𝑥g ∼ 𝑁(0, 𝜎 5/2)
• 𝑥 = 𝑥f + 𝑗𝑥g is circularly symmetric complex
Gaussian distributed,
𝑥 ∼ 𝐶𝑁(0, 𝜎 5), with
d d
1 c i 1 cj
b d b
𝑝R 𝑥 = 𝑒 e 𝑒 ed
𝜋𝜎 5 d 𝜋𝜎 5
1 b cd
= 5 𝑒 e
𝜋𝜎
Properties: 𝐸 𝑥 = 0, 𝐸 𝑥 5 = 𝜎5
TSKS14 Multiple Antenna Communications 2020-03-27 19
Capacity and mutual information
• Channel capacity
𝐶 = max 𝐼(𝑥; 𝑦) 𝑥 Channel 𝑦
lm (c)
• Mutual information: 𝐼 𝑥; 𝑦 = ℎ 𝑦 − ℎ(𝑦|𝑥)
• Differential entropy:
ℎ 𝑦 = −𝐸 log 5 𝑝P 𝑦 ≤ log 5 𝜋𝑒Var 𝑦
• Conditional diff. entropy:
ℎ 𝑦|𝑥 = −𝐸 log 5 𝑝P|R (𝑦|𝑥) Equality if
Complex Gaussian
TSKS14 Multiple Antenna Communications 2020-03-27 20
Differential entropy ℎ 𝑥 of 𝑥 ∼ 𝐶𝑁(0, 𝑝)
• Direct computation
1 b cd 1 b cd
ℎ 𝑥 = −r 𝑒 l log 5 𝑒 l 𝑑𝑥
c∈ℂ 𝜋𝑝 𝜋𝑝
1 b cd 𝑥5
=r 𝑒 l
log 5 𝜋𝑝 + log 5 𝑒 𝑑𝑥
c∈ℂ 𝜋𝑝 𝑝
d cd
1 b cl log 5 𝑒 𝑥5 b l
= log 5 𝜋𝑝 r 𝑒 𝑑𝑥 + r 𝑒 𝑑𝑥
c∈ℂ 𝜋𝑝 𝑝 c∈ℂ 𝜋𝑝
log 5 𝑒
= log 5 𝜋𝑝 ⋅ 1 + 𝐸 𝑥 5 = log 5 𝜋𝑒𝑝
𝑝
TSKS14 Multiple Antenna Communications 2020-03-27 21
Capacity of complex AWGN channel
𝑛 ∼ 𝐶𝑁(0, 𝑁3)
𝑥 + 𝑦 =𝑥+𝑛
• Recall: 𝐼 𝑥; 𝑦 = ℎ 𝑦 − ℎ(𝑦|𝑥)
ℎ 𝑦 𝑥 = 𝑦|𝑥 ∼ 𝐶𝑁(𝑥, 𝑁3) = log 5 𝜋𝑒𝑁3
• Mutual information maximized by 𝑥 ∼ 𝐶𝑁(0, 𝑞)
ℎ 𝑦 ≤ 𝑦 = 𝑥 + 𝑛 ∼ 𝐶𝑁(0, 𝑞 + 𝑁3) = log 5 𝜋𝑒(𝑞 + 𝑁3 )
• Capacity:
𝑞
𝐶 = ℎ 𝑦 − ℎ 𝑦 𝑥 = log 5 1+
𝑁3
TSKS14 Multiple Antenna Communications 2020-03-27 22
Point-to-point scalar channel
• Continuous-time AWGN channel with channel response 𝑔 ∈ ℂ
• Bandwidth 𝐵 (Hz)
• Power 𝑃 (Watt)
• Noise power spectral density 𝑁3 (Watt/Hz)
• Sampling: 2𝐵 real samples/s, 𝐵 complex samples/s:
𝑦 𝑙 =𝑔⋅𝑥 𝑙 +𝑛 𝑙 , 𝑙 = sample index
• Energy per sample: 𝑞 = 𝑃/𝐵
• 𝑥 𝑙 ∼ 𝐶𝑁(0, 𝑃/𝐵)
• 𝑛 𝑙 ∼ 𝐶𝑁(0, 𝑁3) Complex Gaussian signals are capacity achieving
TSKS14 Multiple Antenna Communications 2020-03-27 23
Point-to-point scalar channel: Capacity
• Sampling: 2𝐵 real samples/s, 𝐵 complex samples/s:
𝑦 𝑙 = 𝑔 ⋅ 𝑥 𝑙 + 𝑛[𝑙]
• 𝑥 𝑙 ∼ 𝐶𝑁(0, 𝑃/𝐵)
• 𝑛 𝑙 ∼ 𝐶𝑁(0, 𝑁3)
• Capacity:
• Depends on 𝑃 and 𝐵
|}d
C = log 5 1 + ^~ bits per complex sample
•
|}d
𝐶 = 𝐵 ⋅ log 5 1 + bits per second
^~•
TSKS14 Multiple Antenna Communications 2020-03-27 24
𝑃
Capacity behaviors 𝑁3
log 5 𝑒
8
• SNR:
𝑃 7
𝐵𝑁3 6
Power limited region
5
• log 5 1 + SNR
• Linear for small SNR 4
• Almost flat for large SNR 3 Bandwidth
2 limited
region
1
|}d
Assumption: = 5 ⋅ 10ƒ Hz
~•
0
0 20 40 60 80 100
TSKS14 Multiple Antenna Communications 2020-03-27 25
Summary
• Capacity of point-to-point channel:
𝑃
𝐶 = 𝐵 ⋅ log 5 1+ bits per second
𝐵𝑁3
|
• SNR per degree of freedom (symbol): ^~•
• Behaviors
• Low SNR: 𝐶 grows linearly with power 𝑃
• High SNR: 𝐶 grows linearly with bandwidth 𝐵
End of Lecture 1
TSKS14 Multiple Antenna
Communications