TSKS14

Multiple Antenna
Communications
Lecture 5, 2020

Emil Björnson
TSKS14 Multiple Antenna Communications 2020-04-17 2

Outline of this lecture

• Practical issues with point-to-point MIMO

• Introduction to multi-user MIMO

• Uplink and downlink
• Orthogonal access versus spatial multiplexing

• Capacity region
• Operating points
• Uplink capacity region
TSKS14 Multiple Antenna Communications 2020-04-17 3

Recall: Point-to-Point MIMO Capacity

• Compute SVD of channel matrix:
+
𝑮 = 𝑼𝜮𝑽& = ' 𝑠( 𝒖( 𝒗&
(
()*
• 𝜮 “diagonal” with 𝑠*, … 𝑠+
• 𝑼 = 𝒖* … 𝒖1
• 𝑽 = 𝒗* … 𝒗2

Decompose the channel into

𝑆 parallel channels
TSKS14 Multiple Antenna Communications 2020-04-17 4

Problems with point-to-point MIMO

• Multiplexing gain: 𝑆 = rank 𝑮
• Line-of-sight: 𝑆 ≈ 1
• Non-line-of-sight: 𝑆 = min 𝑀, 𝐾
Mainly beamforming gain:
• High SNR:
Likely to be in line-of-sight (LOS)
• Low SNR:
Likely to be non-line-of-sight (NLOS)

Not scalable:
User devices are small, cannot fit many
antennas
TSKS14 Multiple Antenna Communications 2020-04-17 5

Multiuser MIMO Communication

• Uplink
• From users to base station
• Multipoint-to-point MIMO
Multiantenna
base station
• D0wnlink
• From base station to users
• Point-to-multipoint MIMO
TSKS14 Multiple Antenna Communications 2020-04-17 6

Orthogonal multiple access

• Two users want to communicate with base station 𝛽 𝛽
• Power per user: 𝑃
• Bandwidth: 𝐵
• Noise power spectral density: 𝑁A User 1 User 2

• Divide bandwidth: 𝛼𝐵 to user 1, (1 − 𝛼)𝐵 to user 2

𝑃𝛽
𝑅* = 𝛼𝐵 log J 1 +
𝛼𝐵𝑁A
𝑃𝛽
𝑅J = 1 − 𝛼 𝐵 log J 1 +
1 − 𝛼 𝐵𝑁A
TSKS14 Multiple Antenna Communications 2020-04-17 7

Orthogonal multiple access: Rate region

• Rates depend on 𝛼:
𝑃𝛽
𝑅* = 𝛼𝐵 log J 1 +
𝛼𝐵𝑁A 𝛼=0
𝑃𝛽
𝑅J = 1 − 𝛼 𝐵 log J 1+
1 − 𝛼 𝐵𝑁A

1
𝛼=
2

What is the preferred

operating point? 𝑃𝛽
= 10 dB 𝛼=1
𝐵𝑁A
TSKS14 Multiple Antenna Communications 2020-04-17 8

From user 1 Noise

Non-orthogonal multiple access 𝑥* ~𝐶𝑁(0, 𝑃) (power 𝐵𝑁A )

• Let both users transmit simultaneously: 𝑦 = 𝑥* + 𝑥J + 𝑛

Received signal
From user 2
𝑥J ~𝐶𝑁(0, 𝑃)
• Strategy:
1. Decode signal from user 1, treat interference as noise
𝑃
𝑅* = log J 1 +
𝑃 + 𝐵𝑁A
2. Subtract 𝑥*: 𝑦 − 𝑥* = 𝑥J + 𝑛. Decode signal from user 2:
𝑃
𝑅J = log J 1 +
𝐵𝑁A

Called: Successive interference cancelation We can change the user order

 TSKS14 Multiple Antenna Communications 2020-04-17 9

Non-orthogonal multiple access: Rate region

2
3
Four operating points 𝑅*, 𝑅J :
W
1. log J 1 + XY ,0
Z
W
2. 0, log J 1 + XY
Z
W W
3. log J 1 + W[XY , log J 1 + XY
Z Z
W W
4. log J 1 + XY , log J 1 + W[XY
Z Z 4

Time sharing:
We can achieve all points in between
1
TSKS14 Multiple Antenna Communications 2020-04-17 10

Uplink in Multiuser MIMO

𝑥*
𝑦*
𝑦J
Transmitters 𝑥] Receiver

𝑥2 𝑦1

• Notation:
• 𝐾 single-antenna users, 𝑀 base station antennas
^
• Channel response 𝑔] from user 𝑖 to antenna 𝑗
• Data signals 𝑥*, … , 𝑥2 , received signals 𝑦*, … , 𝑦1
TSKS14 Multiple Antenna Communications 2020-04-17 11

Uplink Multiuser MIMO: System model

• Received signal:
𝒚 = 𝜌cd 𝑮𝒙 + 𝒘
where
𝑦* 𝑔** ⋯ 𝑔2* 𝑥* 𝑤*
𝒚= ⋮ 𝑮= ⋮ ⋱ ⋮ 𝒙= ⋮ 𝒘= ⋮
𝑦1 𝑔*1 ⋯ 𝑔21 𝑥2 𝑤1

• Parameters are normalized: SNR is 𝜌cd

• Each users signal is power-limited as 𝔼 𝑥( J ≤1
• Normalized noise: 𝒘 ∼ 𝐶𝑁(𝟎, 𝑰1 )
TSKS14 Multiple Antenna Communications 2020-04-17 12

What is the difference from point-to-point MIMO?

• Difference 1: Users do not cooperate
• 𝑥*, … , 𝑥2 are independent signals
• Difference 2: Each user cares about its own performance
• 𝐾 user capacities instead of one capacity
• Difference 3: Each user has its own power budget
• Power constraint: 𝐸 𝑥( J ≤ 1
• Difference 4: The channel matrix 𝑮 is modeled differently
• Each column can be modeled as a SIMO channel
 TSKS14 Multiple Antenna Communications 2020-04-17 13

Motivating example
Multi-user MIMO
Orthogonal access Non-orthogonal access 𝑅J
𝑅J 𝑅J

𝑅* 𝑅* 𝑅*

Capacity or rate region? Two benefits:

Capacity region is the largest Beamforming gain for every user
possible rate region Smaller interference loss
TSKS14 Multiple Antenna Communications 2020-04-17 14

Shape of capacity region

• One can pick two points and use them fractions of the time
• Similar to time-sharing
• Hence: Line between any two points are in the region
• Conclusion: Region must be a convex set

Possible Impossible
shape shape
 TSKS14 Multiple Antenna Communications 2020-04-17 15

Points in the capacity region

• Combinations (𝑅*, 𝑅J) of rates that can be simultaneously achieved
𝑅J Max sum rate
Many choices!
(Largest 𝑅* + 𝑅J )
Only rule:
Always pick something
Largest 𝑅J Max-min rate on the outer boundary
(Largest 𝑅* = 𝑅J ) (Pareto boundary)!
Capacity region

Largest 𝑅* Max sum rate

𝑅* Fairness scale:
Max-min
fairness
TSKS14 Multiple Antenna Communications 2020-04-17 16

Sum Capacity of Uplink Multiuser MIMO

• Recall: Received signal:
𝒚 = 𝜌cd 𝑮𝒙 + 𝒘
• Assume a deterministic 𝑮
• Let all users transmit with full power: 𝒙~𝐶𝑁(𝟎, 𝑰2 )

Like a point-to-point MIMO channel

But with a “suboptimal” signal covariance matrix 𝑸 = 𝑰1

• Sum rate: 𝑅* + ⋯ + 𝑅2 = log J det 𝑰1 + 𝜌cd 𝑮𝑮&

This is the sum capacity!
Achieved by successive interference cancellation
Decoding order determines who gets which share
TSKS14 Multiple Antenna Communications 2020-04-17 17

Uplink Capacity Region with 𝐾 = 2

• Region contains all (𝑅*, 𝑅J) satisfying:
𝑅* ≤ log J 1 + 𝜌cd 𝒈* J
𝑅J ≤ log J 1 + 𝜌cd 𝒈J J
𝑅* + 𝑅J ≤ log J det 𝑰1 + 𝜌cd 𝑮𝑮&

𝑅J 2 3
Limited by sum capacity

4
𝑮 = 𝒈* 𝒈J

𝑅*
1
TSKS14 Multiple Antenna Communications 2020-04-17 18

Summary
• Point-to-point MIMO channels
• Large multiplexing gains are hard to achieve in practice

• Multi-user MIMO channels

• Similar system model
• Key differences: Independent users, different power, different performance
• Capacity and rate regions
• Orthogonal and non-orthogonal access
 End of Lecture 5

TSKS14 Multiple Antenna

Communications