34_PR_CHOI
Transcript of 34_PR_CHOI
-
8/6/2019 34_PR_CHOI
1/28
Spherical Linear Interpolation forSpherical Linear Interpolation for
TransmitTransmit BeamformingBeamforming in MIMOin MIMO--OFDMOFDMSystems with Limited FeedbackSystems with Limited Feedback
Jihoon Choi and Robert W. Heath Jr.
Wireless Networking and Communications Group
The University of Texas at Austin
Abstract Presentation Next AbstractPaper
-
8/6/2019 34_PR_CHOI
2/28
Wireless Networking and Communication Group 2
OutlineOutline
Overview
Diversity techniques in MIMO systems
Closed-loop transmit beamforming for MIMO-OFDM
Problem Statement
Proposed Beamforming Scheme Proposed spherical interpolator
Optimization of phase rotation
Beamformer quantization + interpolation
Numerical Simulations and Discussions
Abstract Presentation Back to TOC
Next Abstract
Paper
-
8/6/2019 34_PR_CHOI
3/28
Wireless Networking and Communication Group 3
Transmit Diversity in MIMO SystemsTransmit Diversity in MIMO Systems
No channel state information (CSI) at Transmitter
Space-time codes [Tarokh et al. 1998]
Full CSI at Transmitter
Antenna selection [Wittneben et al.1994], [Heath et al. 1998] Maximum ratio transmission [Lo 1999], [Andersen 2000]
Partial CSI
Adaptive beamforming [Xia et. al. 2004]
Transmitter
MIMO channel
Receiver
-
8/6/2019 34_PR_CHOI
4/28
Wireless Networking and Communication Group 4
BeamformingBeamforming and Combiningand Combining
Covert MIMO to SISO
Transmit beamforming vectorw
Receive combining vectorz
Coding &
Modulation...
Hw..
Detectionand
Decodingz
Feedback Channel
( with limited data rate )
Estimation &
detection
Complex
number
-
8/6/2019 34_PR_CHOI
5/28
Wireless Networking and Communication Group 5
Max SNR Solution (MRT/MRC)Max SNR Solution (MRT/MRC)
SNR is proportional to
Max SNR solution is
Solution is not unique: ifw is optimal, so is wej
Other solutions possible (constrain w)
Selection diversity
Equal gain combining / transmission
-
8/6/2019 34_PR_CHOI
6/28Wireless Networking and Communication Group 6
Beamforming in MIMOBeamforming in MIMO--OFDMOFDM
P/S
&
AddCP
IDFT
)1(s
)(Ns
IDFT
DFT
)1(1y
)(1 Ny
DFT
)1(rM
y
)(NyrM
M M
M
M
M
M
M
)1(1w
)(1 Nw
M)1(tMw
)(NwtM
M MM M
Feedback of Channel
State Information
M
P/S
&
AddCP
Remove
CP
&S/P
Remove
CP
&
S/P
1n
rMn
)1(1z
)(1 Nz
)1(rM
z
)(NzrM
M
)1(r
)(Nr
-
8/6/2019 34_PR_CHOI
7/28Wireless Networking and Communication Group 7
Signal Model for OneSignal Model for One SubcarrierSubcarrier
Equivalent model for subcarrierk
(identical with a narrowband MIMO case))(1,1 kh
)(, kh tr MM
)(1 kw )(1 kz)(ks )(krM
M M
M
)(kwtM
)(kzrM
M
)(1 kn
)(knrM
Noise
-
8/6/2019 34_PR_CHOI
8/28
Wireless Networking and Communication Group 8
Problem SummaryProblem Summary
Beamforming in MIMO-OFDM requires
Feedback requirements Number of subcarriers
How can we reduce the number of vectors fed back?
Exploit correlation between vectors Send back fraction of vectors
Use smart interpolation
How can limit the feedback for each vector?
Use quantized beamforming [Love et. al. 2003]
-
8/6/2019 34_PR_CHOI
9/28
Wireless Networking and Communication Group 9
Correlation Between SubcarriersCorrelation Between Subcarriers
Subchannelcorrelation
Beamformer correlation
Simulations
h(d): M
t=4, M
r=1, N=64, K=8, L={6,12}
w(d): M
t=4, M
r=2, N=64, K=8, L={6,12}
subsampling rate number of channel taps
-
8/6/2019 34_PR_CHOI
10/28
Wireless Networking and Communication Group 10
Clustering of SubcarriersClustering of Subcarriers
Clustering (K=5)
Disadvantages
Performance degradation in cluster boundary
Cluster size (or feedback reduction) is limited
subcarriers
L
cluster 1 cluster 2 cluster 3 cluster 4
feedback L
-
8/6/2019 34_PR_CHOI
11/28
Wireless Networking and Communication Group 11
Proposed Beamforming MethodProposed Beamforming Method
Subsampling of beamforming vectors
LL L
Reconstruction of beamforming vectors
Transmit power constraints force
Spherical interpolation
subcarriers
L
L
feedbackL
-
8/6/2019 34_PR_CHOI
12/28
Wireless Networking and Communication Group 12
Conventional Spherical InterpolationConventional Spherical Interpolation
Spherical averaging [Watson 1983], [Buss et al. 2001]
Nonuniqueness of optimal beamforming vectors
Random phase rotation has significant impact on interpolations
Performance degradation of spherical interpolators
-
8/6/2019 34_PR_CHOI
13/28
Wireless Networking and Communication Group 13
Proposed InterpolatorProposed Interpolator
Optimization of phase rotation parameters Maximize the diversity gain
Maximize the mutual information
Spherical Linear Interpolator with phase rotation
is a parameter for phase rotation.
-
8/6/2019 34_PR_CHOI
14/28
Wireless Networking and Communication Group 14
Optimization of Phase RotationOptimization of Phase Rotation
Maximizing the minimum channel gain (diversity)
Difficult to get a closed-form solution
Grid search
and Pis the number of quantizedlevels
-
8/6/2019 34_PR_CHOI
15/28
Wireless Networking and Communication Group 15
Alternative SolutionAlternative Solution
Observation
Subcarrier (K/2+1) suffers from the largest distortion
Subcarrier (K/2+1) has the worst average channel gain
Approximation of cost function
-
8/6/2019 34_PR_CHOI
16/28
Wireless Networking and Communication Group 16
ClosedClosed--Form SolutionForm Solution
Differentiation with respect to
1 and 2 are complex constants satisfying
Always, there exist two solutions.
Second derivative
The solution making the second derivative be negative is the
optimal solution. closed-form solution
-
8/6/2019 34_PR_CHOI
17/28
Wireless Networking and Communication Group 17
Optimization in Terms of CapacityOptimization in Terms of Capacity
Maximizing the capacity
Grid search
Closed-form solution is the same as in diversity case
-
8/6/2019 34_PR_CHOI
18/28
Wireless Networking and Communication Group 18
Simulation EnvironmentsSimulation Environments
MIMO channels Each channel impulse response has 6 taps (L=6) with uniform
profile and follows i.i.d.CN
(0,1/L
). The channels between different transmit and receiver antennaspairs are independent.
Noise: i.i.d. with CN(0,N0)
Simulation parameters M
t=4 (# of Tx antennas), M
r=2 (# of Rx antennas)
N=64 (# of subcarriers), K=8 (subsampling rate), L=6
Basic assumptions MRC at the receiver
No delay and no transmission error in the feedback channel
QPSK, no water-filling
-
8/6/2019 34_PR_CHOI
19/28
Wireless Networking and Communication Group 19
Simulation ResultsSimulation Results Channel GainChannel Gain
OFDM subcarrier vs. effective channel gain
-
8/6/2019 34_PR_CHOI
20/28
Wireless Networking and Communication Group 20
Simulation ResultsSimulation Results UncodedUncoded BERBER
No channel coding
-
8/6/2019 34_PR_CHOI
21/28
Wireless Networking and Communication Group 21
Simulation ResultsSimulation Results CapacityCapacity
No power control for diversity techniques
-
8/6/2019 34_PR_CHOI
22/28
Wireless Networking and Communication Group 22
Simulation with QuantizationSimulation with Quantization
Quantization of beamforming vectors
By using the codebook in [Love & Heath 2003]
6 bits of feedback for each beamforming vector
Phase optimization
Grid search with P=4 (2 bits of feedback per phase)
Feedback requirements Ideal: 6N= 384 bits / OFDM symbol
Clustering: 6N/K= 48 bits / OFDM symbol
Selection: 2N= 128 bits / OFDM symbol Proposed: 8N/K = 64 bits / OFDM symbol
Orthogonal STBC (OSTBC): none
-
8/6/2019 34_PR_CHOI
23/28
Wireless Networking and Communication Group 23
Simulation ResultsSimulation Results -- Channel GainChannel Gain
Mt=4, Mr=2, N=64, K=8, L=6
-
8/6/2019 34_PR_CHOI
24/28
-
8/6/2019 34_PR_CHOI
25/28
Wireless Networking and Communication Group 25
Simulation ResultsSimulation Results -- BERBER
BER degradation by channel prediction error
Prediction error model
-
8/6/2019 34_PR_CHOI
26/28
Wireless Networking and Communication Group 26
Simulation ResultsSimulation Results -- CapacityCapacity
No power control
-
8/6/2019 34_PR_CHOI
27/28
Wireless Networking and Communication Group 27
Simulation ResultsSimulation Results Coded BERCoded BER
1/2 convolutional code
Interleaving/deinterleaving
Frame length
30 OFDM symbols 64 bits / OFDM symbol
Abstract Presentation Back to TOCPaper
-
8/6/2019 34_PR_CHOI
28/28
Wireless Networking and Communication Group 28
ConclusionsConclusions
Summary
Interpolation for transmit beamforming in MIMO-
OFDM proposed
Relates to problem of spherical quantization
Future work
Proposed system does not obtain MIMO capacity
Develop interpolated waterfilling solutions
Next Abstract
p