What is the Role of MIMO in
Future Cellular Networks:
Massive? Coordinated?
mmWave?
Robert W. Heath Jr.
The University of Texas at Austin
Wireless Networking and Communications Group
www.profheath.org
Presentation (c) Robert W. Heath Jr. 2013
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
2
MIMO in Cellular Systems
# of parallel data streams is the
multiplexing gain
2 data streams
Point-to-point MIMO
2 data streams
limited by # user antennas
sum of the # of parallel data
streams is the multiplexing gain
2 data streams
Multiuser MIMO
limited by # BS antennas
(c) Robert W. Heath Jr. 2013
3
Where is MIMO Headed?
Coordinated MIMO
Massive MIMO
B
B
B
mmWave MIMO
Candidate architectures for 5G cellular
(c) Robert W. Heath Jr. 2013
4
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
5
What is Network MIMO?
Out-of-cell
interference
cellular backhaul network
Coordinated transmission from multiple base stations
Known as CoMP or Cooperative MIMO or base station coordination
Interference turned from foe to friend
Exploits presence of a good backhaul connection
Used to improve area spectral efficiency, system capacity
(c) Robert W. Heath Jr. 2013
6
Network MIMO Architectures
eNodeB
eNodeB
eNodeB
Like DAS uses local
coordination
Coordinate over backhaul
Cloud RAN
Coordination clusters
Dynamic coordination
Processing for many base
stations using cloud
(c) Robert W. Heath Jr. 2013
7
Potential Gains from Coordination
1500
1000
90
19 cells and 3 sectors per cell
BS-MS distance= 192 m
ISD=500m
BS
MS
80
0
−500
sum rates per cell
70
Sum-rates (bits/sec/Hz)
500
K=1
K=3
K=9
60
Linear increase
50
40
Unbounded gains
30
20
−1000
−1500
−1500
10
−1000
−500
0
500
1000
1500
0
0
Grid model for a cellular network
5
10
15
SNR (dB)
20
25
30
Throughput gains when out-of-cluster interference is ignored
More cooperation leads to higher gains
Cell edge pushed further out, no uncoordinated interference in the cell
(c) Robert W. Heath Jr. 2013
8
Addressing Out-of-Cell Interference
1500
1000
19 cells and 3 sectors per cell
BS-MS distance= 192 m
ISD=500m
25
BS
MS
20
Sum-rates (bits/sec/Hz)
500
0
−500
K=1
K=3
K=9
15
Bounded gains
10
Sum-rates saturation
5
−1000
−1500
−1500
0
−1000
−500
0
500
1000
1500
0
5
10
15
SNR (dB)
20
25
30
Grid model for a cellular network
Performance saturates with out-of-cluster interference
30% performance gains observed in industrial settings
Be mindful of the saturation point
A. Lozano, R. W. Heath, Jr., and J. G. Andrews, ``Fundamental Limits of Cooperation" to appear in
the IEEE Trans. on Info. Theory. Available on ArXiv.
(c) Robert W. Heath Jr. 2013
9
Critical Issues with Network MIMO
Out-of-cell interference
channel state feedback overhead
When included, results are not as good
Feedback overhead
Need channel state information
Performance seriously degrades
eNodeB
Control channel overhead
eNodeB
Backhaul delay
Increased reference signal overhead
Backhaul link constraints
Reference signal
Data payload
Backhaul link latency (delayed CSI and data sharing)
Cluster edge effects
Coordination still has a cell edge with fixed clusters
Dynamic clustering solves the problem, but more more implementation overhead
(c) Robert W. Heath Jr. 2013
10
Network MIMO Conclusions
Observations
Network MIMO promises a way to get rid of interference
...Yet uncoordinated interference still limits high SNR performance
Backhaul constraints and system overheads further reduce performance gains
General disconnect between academia and industry on the potential
Forecast
Already incorporated into 4G, coordination will be part of 5G as well
Architectures will evolve to support network MIMO-like coordination
•
•
•
Distributed antenna systems are a good starting point
Distributed radio access networks are a likely evolution point
Cloud radio access networks are a possible end objective
(c) Robert W. Heath Jr. 2013
11
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
12
What is Massive MIMO?
hundreds of BS antennas
tens of users
A very large antenna array at each base station
An order of magnitude more antenna elements in conventional systems
A large number of users are served simultaneously
An excess of base station (BS) antennas
Essentially multiuser MIMO with lots of base station antennas
T. L. Marzetta, “Noncooperative cellular wireless with unlimited numbers of base station antennas,” IEEE Trans. Wireless Commun., vol. 9, no. 11, pp.
3590–3600, Nov. 2010.
(c) Robert W. Heath Jr. 2013
13
Massive MIMO Key Features
Uplink
training
pilot
contamination
Benefits from the (many) excess antennas
Simplified multiuser processing
Reduced transmit power
Thermal noise and fast fading vanish
Differences with MU MIMO in conventional cellular systems
Time division duplexing used to enable channel estimation
Pilot contamination limits performance
F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta, O. Edfors, and F. Tufvesson, “Scaling up MIMO: Opportunities and challenges with very
(c) Robert W. Heath Jr. 2013
large arrays,” IEEE Signal Processing Mag., vol. 30, no. 1, pp. 40–60, Jan. 2013.
14
Massive MIMO Key Features
Downlink
channel
reciprocity
interference
asymptotic
orthogonality
Benefits from the (many) excess antennas
Simplified multiuser processing
Reduced transmit power
Thermal noise and fast fading vanish
Differences with MU MIMO in conventional cellular systems
Time division duplexing used to enable channel estimation
Pilot contamination limits performance
F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta, O. Edfors, and F. Tufvesson, “Scaling up MIMO: Opportunities and challenges with very
(c) Robert W. Heath Jr. 2013
large arrays,” IEEE Signal Processing Mag., vol. 30, no. 1, pp. 40–60, Jan. 2013.
14
Centralized vs. Distributed
Centralized
7 cells without sectorization
12 users uniformly distributed in each cell
ISD = 500m
Distributed
BS antenna
clusters
BS
Fixed number of base station antennas per cell
* K. T. Truong and R. W. Heath, Jr., “Impact of Spatial Correlation and Distributed Antennas for Massive MIMO systems,” to
appear in the Proceedings of the Asilomar Conference on Signals, Systems, and Computers, Nov. 3-6, 2013.
(c) Robert W. Heath Jr. 2013
15
Potential Gains from Massive MIMO
Uplink
Downlink
60
45
55
40
50
35
number of
number of
45
30
40
25
35
30
20
25
15
10
24
20
48
72
96
120
144
Number of base station antennas
168
15
24
48
72
96
120
144
Number of base station antennas
168
7 cells without sectorization, 12 users uniformly distributed in each cell, ISD = 500m
Distributing antennas achieves higher gains
Saturation is not observed without huge # of antennas
(c) Robert W. Heath Jr. 2013
16
Critical Issues in Massive MIMO
Gains are not that big with not-so-many antennas
Require many antennas to remove interference
Need more coordination to remove effects of pilot contamination
Massive MIMO seems to be more “uplink driven”
Certain important roles are reserved between base stations and users
A different layout of control and data channels may be required
Practical effects are not well investigated
Channel aging affects energy-focusing ability of narrow beams
Spatial correlation reduces effective DoFs as increasing number of antennas
Role of asynchronism in pilot contamination and resulting performance
(c) Robert W. Heath Jr. 2013
17
Massive MIMO Conclusions
Observations
Pilot contamination is a big deal, but possibly overcome by coordination
Performance is sensitive to channel aging effects *
Good performance can be achieved with distributed antennas *
Not clear how to pack so many microwave antennas on a base station
Needs more extensive simulation study with realistic system parameters
Forecast
Massive MIMO will probably not be used in isolation
Will be combined with distributed antennas or base station coordination
•
•
Reduces the effects of pilot contamination
Work with smaller numbers of antennas
* K. T. Truong and R. W. Heath, Jr., “Effects of Channel Aging in Massive MIMO Systems,” to appear in the Journal of
Communications and Networks, Special Issue on Massive MIMO, February 2013.
(c) Robert W. Heath Jr. 2013
18
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
19
Why mmWave for Cellular?
1G-4G cellular
Microwave
300 MHz
3 GHz
5G cellular
millimeter wave
28 GHz
38-49 GHz
70-90 GHz
300 GHz
Cellular systems live with a little microwave spectrum
600MHz total (best case) in the US
Spectrum efficiency is king (MIMO, MU MIMO, HetNets)
Huge amount of spectrum available in mmWave bands [KhanPi]
29GHz possibly available in 23GHz, LMDS, 38, 40, 46, 47, 49, and E-band
mmWave already used in LAN, PAN, and VANET
mmWave links are used for backhaul in cellular networks
(c) Robert W. Heath Jr. 2013
20
Antenna Arrays are Important
highly directional MIMO transmission
Baseband
Processing
Baseband
Processing
antennas are small (mm)
~100 antennas
used at TX and RX
New challenges faced by mmWave cellular
Larger bandwidth means higher noise power and lower SNR
Smaller wavelength means smaller captured energy in an antenna
Solution: Exploit array gain from large arrays
(c) Robert W. Heath Jr. 2013
21
Coverage Gains from Large Arrays
I. M EASUREMENT R ESULTS
LOS & blocked path loss model from measurements.
A Poisson layout of mmWave BSs.
Average cell radius Rc=100m.
A Boolean scheme building model.
1
0.9
Serving BS
Typical User
Buildings
Coverage Probability
0.8
0.7
0.6
Gain from directional antenna array
0.5
0.4
Omni Directional
0.3
PPP Interfering BSs
o
Directional: HPBW=40
Directional: HPBW=10o
0.2
−10
−8
−6
−4
−2
0
2
4
6
8
10
SINR Threshold in dB
Fig. 1.
mmWave networks
can provide acceptable coverage
Directional array gain compensates severe path loss
1
Smaller beamwidth reduces the effect of interference
0.9
0.8
(c) Robert W. Heath Jr. 2013
22
Critical Issues in mmWave MIMO
Dealing with hardware constraints
Need a combination of analog and digital beamforming
Array geometry may be unknown, may change
Performance in complex propagation environments
Evaluate performance with line-of-sight and blocked signal paths
Must adapt to frequent blockages and support mobility
Entire system must support directionality
Need approval to employ the spectrum
(c) Robert W. Heath Jr. 2013
23
mmWave Conclusions
Observations
Coverage may be acceptable with the right system configuration
Strong candidate for higher per-link data rates
Hardware can leverage insights from 60GHz LAN and PAN
Highly directional antennas may radically change system design
Supporting mobility may be a challenge
Forecast
Will be part of 5G if access to new spectrum becomes viable
Most likely will co-exist with microwave cellular systems
Will remain useful for niche applications like backhaul
(c) Robert W. Heath Jr. 2013
24
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
25
Stochastic Geometry for Cellular
performance
analyzed for a
typical user
base station locations
distributed (usually) as a
Poisson point process (PPP)
Stochastic geometry is a useful tool for analyzing cellular nets
Reasonable fit with real deployments
Closed form solutions for coverage probability available
Provides a system-wide performance characterization
J. G. Andrews, F. Baccelli, and R. K. Ganti, "A Tractable Approach to Coverage and Rate in Cellular Networks", IEEE Transactions on Communications, November 2011.
T. X. Brown, "Cellular performance bounds via shotgun cellular systems," IEEE JSAC, vol.18, no.11, pp.2443,2455, Nov. 2000.
(c) Robert W. Heath Jr. 2013
26
Comparing Different Approaches
CoMP model
Sectored cooperation model
Typical user can be edge or center user of the cluster
Several assumptions made to permit calculation
mmWave model
Directional antennas are incorporated as marks of the base station PPP
Blockages due to buildings incorporated via random shape theory
Massive MIMO model
Analyze asymptotic case with infinite number of antennas at the base station
No spatial correlation, includes estimation error, pilot contamination
New expressions derived for each case
Tianyang Bai and R. W. Heath, Jr., `` Asymptotic Coverage Probability and Rate in Massive MIMO Networks ,'' submitted to IEEE Wireless Communications Letters,
May 2013. Available on ArXiv.
(c) Robert W. Heath Jr. 2013
27
SINR Coverage Comparison
Gain from directional
antennas and blockages
in mmWave
1
0.9
SINR Coverage Probability
0.8
0.7
0.6
0.5
Gain from larger number
of antennas
0.4
CoMP: Nt=4, 2 Users, 3 BSs/ Cluster
0.3
Massive MIMO: Nt=∞
mmWave: Nt=64, Rc=100m
0.2
SU MIMO: 4X4
0.1
−10
−5
0
5
SINR Threhold in dB
10
15
20
Fig. 1.
(c) Robert W. Heath Jr. 2013
28
Rate Comparison
1
Gain from larger bandwidth
0.9
0.8
CoMP: Nt=4, 2 Users, 3 BSs/ Cluster
Massive MIMO: Nt=∞
Rate Coverage Probability
0.7
mmWave: Nt=64, Rc=100m
SU MIMO: 4X4
0.6
0.5
0.4
Gain from
serving multiple users
0.3
0.2
0.1
0
0
Fig. 3.
500
1000
1500
2000
2500
3000
Cell Throughput in Mbps
3500
4000
4500
5000
(c) Robert W. Heath Jr. 2013
29
Rate Comparison
Massive
MIMO
MMwave
CoMP
SU MIMO
Signal BW
50 MHz
500 MHz
50 MHz
50 MHz
User/ Cell
8
1
2
1
bps/Hz per user 5.47 bps/Hz
Rate/Cell
Capacity/ Cell
8.00 bps/Hz 4.34 bps/Hz 4.95 bps/Hz
21.88 bps/Hz 8.00 bps/Hz 8.68 bps/Hz 4.95 bps/Hz
1.10 Gbps
4.00 Gbps
434 Mbps
248 Mbps
mmWave outperforms due to more available BW
* of course various parameters can be further optimized
** SU MIMO is 4x4 w/ zero-forcing receiver
(c) Robert W. Heath Jr. 2013
30
Outline
MIMO in cellular networks
Coordinated Multipoint a.k.a. network MIMO
Massive MIMO
Millimeter wave MIMO
Comparison between technologies
Parting thoughts
31
Parting Thoughts
Conclusions
cooperation will be used in some form, more powerful
cooperative
with better infrastructure, need to be mindful of overheads
MIMO
in system design
massive
MIMO
some potential for system rates, need large base station
arrays, can be used with cooperation
mmWave
MIMO
large potential for peak rates, more hardware challenges,
requires more spectrum, more radical system design
potential
(c) Robert W. Heath Jr. 2013
32
Questions?
Robert W. Heath Jr.
The University of Texas at Austin
www.profheath.org