FTW Telekommunikationsforum

June 18th 2010, Vienna, Austria

Antenna selection in MIMO systems

Andreas F. Molisch
Wireless Devices and Systems (WiDeS) Group
University of Southern California (USC)

Antenna selection

2010 A. F. Molisch

Contents

Introduction and motivation

Performance analysis
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

2010 A. F. Molisch

Antenna Selection
Additional costs for MIMO
1. more antenna elements (cheap)
2. more signal processing (Moores law)
3. one RF chain for each antenna element

Basic idea of antenna selection:

have many antenna elements, but select only best for
downconversion and processing
only at one link end: cost reductions might be more
important at one link end (MS) than the other

Hybrid antenna selection: select best L out of

available N antenna elements, use those for
processing

Antenna selection

2010 A. F. Molisch

Transmit antenna selection

DAC

PA

DAC

PA

DAC

PA

S/W

Fewer expensive RF
chains

Antenna selection can be applied at transmitter or

receiver or both

Antenna selection

2010 A. F. Molisch

Contents

Introduction and motivation

Performance analysis
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

2010 A. F. Molisch

Antenna Selection for Diversity

Weight selection if all antenna elements are used

Write channel matrix as H WV H

Excite channel with Vi, receive with WiH

Received power is i2
Antenna and weight selection for H-S/MRT

Create submatrices by striking rows

Compute maximum singular value for this submatrix

Search submatrix that gives largest max. singular val.

Use singular values associated with selected

submatrix as antenna weights
maxSH
Antenna selection

2
maxi i
6

2010 A. F. Molisch

Bounds for the SNR distribution

Upper and lower bounds

1
~2
~2
~2
i max i i

i
min Lt , N r i
i
Determine

bound max
~
SH

~2
i max
SH~

~ 2
h ij

bound

p i 1 , 2 , N t

N!
i1

1
N r 1
N r i

exp i

Antenna selection

for

1 2 N t

otherwise

i 1

where (i) are ordered SNRs with distribution

Nt

Lt

2010 A. F. Molisch

Bounds for the SNR distribution

Characteristic function:

Nt !
j
N r Nt
1

0 d1 N11 e
r

e jL t1 1

d2 N2r 1 e 2 e jL t2 2

Nt1

Nt jL t N t Nt
dNt NNr 1
e
e

Analytical evaluation: recursive algorithm

Antenna selection

2010 A. F. Molisch

HS-MRT vs. MRT

Hybrid selection
cdf(C)

1
0.8

0.6
0.4

0.2

Lt=1
2
3
4

Lt=5
6
7
8
15

10

capacity C [bits/s/Hz]

Maximum ratio transmission

cdf(C)

1
0.8
0.6
0.4
0.2
5

Antenna selection

Nt=1
2
3
4

Nt=5
6
7
8
15

10

capacity C [bits/s/Hz]

2010 A. F. Molisch

Capacity with RX antenna selection

cdf (C)

L=2

L=3

0.5

L=8

0
10

20

30

capacity C [bits/s/Hz]

3 transmit antennas, 20dB SNR

Admissible number of spatial streams: determined by number of RF chains
Slope of capacity cdf: determined by number of antenna elements
Antenna selection

10

2010 A. F. Molisch

Contents

System model
Performance analysis
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

11

2010 A. F. Molisch

Channel Estimation Error Diversity

cdf (C)

1
noisy antenna
selection
noisy TX weights
noisy RX weights

0.5

noisy antenna
selection

noisy antenna
selection
noisy TX weights

ideal channel
channel
ideal
estimation
estimation

0
7

10

11

capacity C [bits/s/Hz]
Antenna selection

12

2010 A. F. Molisch

Channel Correlation - Diversity

Capacity [bits/s/Hz]

10
8/8 system

9.5
9
8.5

3/8 system
with power
selection

3/8 system
with optimum
selection

7.5
7

-2

-1.5

-1

-0.5

0.5

1.5

Log(Lcorr/)
Antenna selection

13

2010 A. F. Molisch

Contents

System model
Performance analysis
Antenna selection algorithms
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

14

2010 A. F. Molisch

Hardware constraints on training

Step 1

Step 1
Rx1

Tx1

Rx1

Rx1
Tx1

Tx1

RF chain

Rx2
Tx2

Rx2
Tx2

Rx2
Tx2

Step 2

Step 2
Rx1

Tx1

Tx1

RF chain

Rx1

Rx1
Tx1

Rx2

Rx2

Tx2

Transmit antenna selection

Rx2

Receive antenna selection

Need to learn channel states of both antennas. But, only 1 antenna

can transmit/receive at any time
Increased time and (pilot) energy overhead
Antenna selection

15

2010 A. F. Molisch

AS training in 802.11n WLANs

Training done using MAC headers only

Consecutive packets train disjoint antenna subsets
Feedback: CQI or antenna indices directly
SIFS

Tx1

AS train1

TXASI

Tx2

AS train2
Data from best antenna

AS train3

Tx4

Rx

SIFS

ASFB

TXASR

Request AS training
Antenna selection

Rx estimates channels
16

AS feedback
2010 A. F. Molisch

Impact of outdated training

Outdated CSI leads to error floor

Analytical consideration: 1-out-of-N selection

Antenna selection

17

2010 A. F. Molisch

Estimation, selection, reception

Antenna selection

18

2010 A. F. Molisch

Closed-form expression of SEP

Antenna selection

19

2010 A. F. Molisch

Optimal weights

Antenna selection

20

2010 A. F. Molisch

Simulation results

Antenna selection

21

2010 A. F. Molisch

Contents

System model
Performance analysis
Antenna selection algorithms
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

22

2010 A. F. Molisch

Principle

Demodulator

RF
1
2 Pre.
.
.
.

Nr

Proc
(M)

S
W
I
T
C
H

1
LNA

LNA

Down
Conv
Down
Conv

Baseband

A/D

Sig
Proc
A/D

Diversity gain AND Beamforming gain maintained

Beam selection Vs Antenna Selection
Can be implemented using variable phase-shifters
In case M is of size L x Nr, selection switch/algorithm is not required

Antenna selection

23

2010 A. F. Molisch

RF Pre-processing Solutions

Channel-independent solution
Fixed matrix (FFT Butler matrix)

Time-variant solution
Elements of pre-processing matrix tuned to instantaneous channel
state

Time-invariant solution
Elements of pre-processing matrix based only on channelstatistics
Can be implemented with or without selection

Antenna selection

24

2010 A. F. Molisch

Channel-independent solution
Transformation matrix is DFT matrix

1
r

e j r

e jr1 r

1 e jr1 r e jr1

Transforms from antenna space to beamspace

Each output of DFT has full beamforming gain N

Antenna selection

25

2010 A. F. Molisch

Time-variant solution
(instantaneous CSI)
For diversity:
Achievable performance: same as with full-complexity CSI
Phase-shifter only solution: optimum performance with 2 RF
chains

Frequent channel sounding required

Antenna selection

26

2010 A. F. Molisch

Time-invariant preprocessing
Time-variant selection
TI

max EH max
w S L MNr MN r S L
MNr
S L Nt

1/ 2

S L MNr Hv

Difficult to handle analytically

SL that depends on instantaneous channel state

Tractable lower bound: Swap max and expectation (EH)

TI max max EH w S L MNr MN r S L
SL
MNr
Nt

Antenna selection

27

1/ 2

S L MNr Hv

2010 A. F. Molisch

Solution that Improves Lower

Bound
Given SL, problem is similar to the LxNr case
Specifies only L columns of M
Remaining (Nr - L) columns must be orthogonal to these L
columns

Given SL, fix corresponding L columns

Subsequent manipulations should not deteriorate
previously considered selections
Successive improvement of lower bound by fixing
remaining columns

MTI P 1 ,, L ,, N r

Antenna selection

28

2010 A. F. Molisch

Beam Patterns

Time-Invariant
(Mean AoA = 450)

FFT Butler

TI adapts to mean AoA and angle spread (unlike FFT)

Adapts to presence of multiple clusters
Antenna selection

29

2010 A. F. Molisch

Receivers with L demodulators

CDF of SNR

Nr = Nt = 4, L = 1
Angle spread = 60
dt/ = 0.5

SNR (dB)

Gain of over 5.0 dB over antenna selection

Mean AoA = 450: Gain of 2.2 dB over FFT-Selection
Mean AoA = 450: No gain over FFT-Selection

Antenna selection

30

2010 A. F. Molisch

Effect of Spatial Correlation

CDF of SNR

Nr = Nt = 4, L = 1
Mean AoA = 600
dt/ = 0.5

SNR (dB)

Time-invariant solution efficacy decreases as correlation decreases

Antenna selection efficacy improves as correlation decreases

Antenna selection

31

2010 A. F. Molisch

Contents

System model
Performance analysis
Antenna selection algorithms
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

32

2010 A. F. Molisch

Measurement setup
Measure channel for Personal Area Networks
Use access point, PC, and handheld

Antenna selection

33

2010 A. F. Molisch

Diversity gain

Antenna selection

34

2010 A. F. Molisch

Impact of antenna configurations

Polarization

Line

Horizontal (H)

Vertical (V)

Saw
lamda/2

10

Empirical CDF

Configurations

10

Alternate H & V (Alt HV)

Rectangular

10

d
10

10

-2

10
H - pol
V - pol
H alt V pol
DP

-4

-5

10

10

Rectangle

10

Saw

-2

H - pol
V - pol
H alt V pol
DP

-4

-5
0

10

Separation

Dual polarized (DP)

Empirical CDF

lamda/2

Line

10

10

-2

10
H - pol
V - pol
H alt V pol
DP

-4

-5
0
5
Normalized SNR [dB]

10

10

-2

-4

DP rect /2
DP rect
DP rect 2
DP line

-5
0
5
Normalized SNR [dB]

10

Configuration comparison for the AP - PC

scenario HS-B at PC only. LOS. 4:22:2.

Antenna selection

35

2010 A. F. Molisch

Contents

System model
Performance analysis
Antenna selection algorithms
Effect of nonidealities
Training
RF preprocessing
Results in measured channels
Hardware aspects
Summary and conclusions

Antenna selection

36

2010 A. F. Molisch

Hardware Aspects
Attenuation of switches
either decrease the effective SNR,
or LNA has to be before switchrequires more LNAs

Switching time: has to be much smaller than duration

of training sequence
Accuracy of switch: transfer function has to be same
from each input to each output port
MEMS switches: have low insertion loss (0.1dB), but
large switching time (5 microsec)
Solid-state switches: high insertion loss (>1 dB), but
short switching times (100 ns)

Antenna selection

37

2010 A. F. Molisch

Phase Quantization Errors

3-bit phase quantization (steps of 450): capacity within 0.1 bits/sec/Hz

2-bit phase quantization (steps of 900): capacity within 0.3 bits/sec/Hz
Spatial diversity: 1 dB loss in mean SNR observed
Antenna selection

38

2010 A. F. Molisch

Summary and Conclusions

Antenna selection retains the diversity degree, but SNR penalty
For spatial multiplexing, comparable capacity if LrNt
Optimum selection algorithms have complexity N!/(N-L)!;
however, fast, good selection algorithms exist
For low-rank channels, transmit antenna selection can increase
capacity
Channel estimation errors do not decrease capacity significantly
Frequency selectivity reduces effectiveness of antenna selection
RF preprocessing greatly improves performance, especially in
correlated channels
Covariance-based preprocessing especially suitable for
frequency-selective channels
Switches with low attenuation required both for TX and RX

Antenna selection

39

2010 A. F. Molisch

Acknowledgements
This work was done in collaboration with

Moe Z. Win
Jack H. Winters
Yang-Seok Choi
Xinying Zhang
Sun-Yuan Kung
Jin-Yun Zhang
Neelesh B. Mehta
Pallav Sudarshan
Yabo Li
Peter Almers

Antenna selection

40

2010 A. F. Molisch

References (1)
Book chapters and journal papers

V. Kristem, N. B. Mehta, and A. F. Molisch, Optimal Receive Antenna Selection in Time-Varying Fading Channels
with Practical Training Constraints, IEEE Trans. Comm., submitted (first round of reviews).
T. Wang, J. Tao, A. F. Molisch, P. Orlik, and J. Zhang, Adaptive Antenna Selection at Mobile Stations for SDMA in
WiMAX Networks, Wireless Computing and Mobile Communications, in press
P. Almers, T. Santos, F. Tufvesson, A. F. Molisch, J. Karedal, and A. Johansson, Antenna selection in measured
indoor channels, IET Microwaves, Antennas & Popagation 1, 1092-1100 (2007).
Y. Li, N. B. Mehta, A. F. Molisch, and J. Zhang, Optimal Signaling and Selection Verification for Transmit Antenna
Selection, IEEE Trans. Comm, 55, 778-789, 2007.
N. Mehta and A. F. Molisch, Antenna selection, in G. Tsoulos (ed.), MIMO antenna technology for wireless
communications, invited, Wiley, (2006).
H. Zhang, A. F. Molisch, and J. Zhang, Applying Antenna Selection in WLANs for Achieving Broadband Multimedia
Communications, IEEE Trans. Broadcasting, 52, 475-482 (2006).
P. Sudarshan, N. B. Mehta, A. F. Molisch, and J. Zhang, Channel Statistics-Based Joint RF-Baseband Design for
Antenna Selection for Spatial Multiplexing, IEEE Trans. Wireless Comm. 5, 3501-3511, (2006)
A. F. Molisch, M. Z. Win, Y. S. Choi, and J. H. Winters, Capacity of MIMO systems with antenna selection, IEEE
Trans. Wireless Comm., 4, 142-154 (2005).
X. Zhang, A. F. Molisch, and S. Y. Kung, Variable-phase-shift-based RF-baseband codesign for MIMO antenna
selection, IEEE Trans. Signal Proc., 53, 4091-4103 (2005).
A. F. Molisch and X. Zhang, FFT-based Hybrid Antenna Selection Schemes for spatially correlated MIMO channels,
IEEE Comm. Lett., 8, 36-38 (2004).
A. F. Molisch and M. Z. Win, MIMO systems with antenna selection, IEEE Microwave Magazine March 2004, 46-56
(2004).
A. F. Molisch, M. Z. Win, and J. H. Winters, Reduced-complexity transmit/receive diversity systems, IEEE Proc.
Signal Proc, special issue on MIMO., 51, 2729-2738 (2003).

Antenna selection

41

2010 A. F. Molisch

References (2)
Conference papers

V. Kristem, N. B. Mehta, and A. F. Molisch, A Novel Energy-Efficient Training Method for Receive Antenna Selection, IEEE ICC 2010, in
press.
V. Kristem, N. B. Mehta, and A. F. Molisch, Optimal Weighted Antenna Selection For Imperfect Channel Knowledge From Training, IEEE
ICC 2009.
T. Wang, Z. Tao, A. F. Molisch, P. Orlik, and J. Zhang, Antenna Selection for SDMA in Next Generation WiMAX Networks, IEEE IWCMC
2009.
C. Nie, Z. Tao, N. B. Mehta, A. F. Molisch, J. Zhang, T. Kuze, and S. Panwar, Antenna Selection for Next Generation IEEE 802.16 Mobile
Stations, IEEE ICC 2008 3457-3462 (2008).
N. B. Mehta, A. F. Molisch, J. Zhang, and E. Bala, Antenna Selection Training in MIMO-OFDM/OFDMA Cellular Systems, IEEE CAMSAP
2007, invited, 2007.
Y. Li, N. B. Mehta, A. F. Molisch, and J. Zhang, Optimal Signaling for Single Transmit Antenna Selection with Erroneous Feedback, Proc.
IEEE Globecom 2006, (2006).
P. Almers, T. Santos, F. Tufvesson, A. F. Molisch, J. Karedal, and A. J. Johansson, Measured Diversity Gains from MIMO Antenna
Selection, Proc. IEEE VTC 2006 fall 1-6 (2006)
H. Zhang, A. F. Molisch, D. Gu, D. Wang, and J. Zhang, Implementing Antenna Selection in Wireless LAN, IEEE International Symposium
on Broadband Multimedia Systems and Broadcasting 2006, Las Vegas, NV, April 6-7, 2006.
P. Sudarshan, N. B. Mehta, A. F. Molisch, and J. Zhang, Channel Statistics-Based Joint RF-Baseband Design for Antenna Selection for
Spatial Multiplexing, Proc. IEEE Globecom 2004, 3947 3951 (2004).
P. Sudarshan, N. B. Mehta, A. F. Molisch, and J. Zhang, Spatial Diversity and Channel Statistics-Based RF-Baseband Co-design for Antenna
Selection, Proc. 60th IEEE Vehicular Techn. Conf.,, invited paper, 1658 - 1662 (2004).
P. Sudarshan, N. B. Mehta, A. F. Molisch, and J. Zhang, Antenna Selection with RF Pre-Processing: Robustness to RF- and Selection- NonIdealities, Proc. IEEE Radio and Wireless Conf. (RAWCON) 2004, 391 394 (2004).
X. Zhang, A. F. Molisch, and S. Y. Kung, Phase-shift-based antenna selection for MIMO channels, Proc. IEEE Globecom 2003, 1089-1093
(2003).
Y. S. Choi, A. F. Molisch, M. Z. Win, and J. H. Winters, Fast antenna selection algorithms for MIMO systems, (invited) Proc. 58th IEEE
Vehicular Techn. Conf., 1733 - 1737 (2003).
A. F. Molisch, X. Zhang, S. Y. Kung, and J. Zhang, FFT-based Hybrid Antenna Selection Schemes for spatially correlated MIMO channels,
Proc. IEEE Symp. Personal Indoor Mobile Radio Comm 2003, 1119-1123 (2003).
A. F. Molisch, MIMO systems with antenna selection an overview, (invited), Proc. IEEE Radio and Wireless Conf. (RAWCON) 2003, 167170 (2003).
A. F. Molisch, M. Z. Win, and J. H. Winters, Performance of Reduced-Complexity Transmit/Receive-Diversity Systems, Proc. IEEE Int. Conf.
Wireless Personal Multimedia Comm. 2002, 738-742 (2002).
A. F. Molisch, M. Z. Win, and J. H. Winters, Reduced-Complexity Transmit/Receive-Diversity Systems , Proc. 53rd IEEE Vehicular Techn.
Conf., 1996-2000 (2001).
A. F. Molisch, M. Z. Win, and J. H. Winters, Capacity of MIMO systems with antenna selection , Proc. IEEE ICC 2001, 570-574 (2001).

Antenna selection

42

2010 A. F. Molisch

Acknowledgement: some slides in this presentation are based on cooperations with

Dr. N. B. Mehta (IISc), Dr. J. Zhang (MERL), Dr. F. Tufvesson (Lund Univ.),
Dr. H. Zhang (Marvell), Dr. X. Zhang (Princeton)

Contact information

Andreas F. Molisch
Ph.D., FIEEE, FIET
Head, Wireless Devices and Systems (WiDeS) Group
Communications Sciences Institute,
Dpt. Of Electrical Engineering
Viterbi School of Engineering
University of Southern California (USC)
Los Angeles, CA, USA

Email: Andreas.Molisch@ieee.org
Website: wides.usc.edu

Antenna selection

43

2010 A. F. Molisch