Transcript Slide 1

Energy-Efficient, Large-scale
Distributed Antenna System
(L-DAS)
under revision for JSTSP
Parts of this work have been presented at the IEEE
GLOBECOM, Atlanta, GA, USA, Dec. 2013
Jingon Joung, Yeow Khiang Chia, Sumei Sun
Modulation and Coding Department
Institute for Infocomm Research, A*STAR
Internal Meeting with Prof. Moe Z. Win
14 January 2014
Motivation
• To achieve high spectral efficiency (SE) and
energy efficiency (EE)
• For high SE
– MU-MIMO: LTE-A beyond Re-7
– Distributed systems: e.g., coordinated multi-point
operation (CoMP), LTE-A Re-11
– Massive (large) MIMO: recent trend
• For high EE
– Power control (PC): efficient-power transmission
L-DAS System
BBU: baseband unit
(signal processing center)
IAD: intra-ant distance
U users
M antennas
H: U-by-M MU-MIMO ch. matrix
S: M-by-U binary AS matrix
W: M-by-U precoding matrix
P: U-dim diagonal PC matrix
x: U-by-1 symbol vector
n: U-by-1 AWGN vector
Objectives & Contribution
•
•
•
•
Study an L-DAS
Provide a practical power consumption model
Formulate an EE maximization problem
Resolve issue on huge signaling, complexity
requirement:
–
–
–
–
Antenna selection (AS) method
Threshold-based user-clustering method
MU-MIMO precoding method
Optimal and heuristic power control methods
• Verify the EE merit of L-DAS
Power Consumption Model
Power consumption
TPI (transmit power independent) term
TPD (transmit power dependent) term
eRF (electric RF)
oRF (optical RF)
Cont.
• TPD term
• TPI term
Pcc1: eRF
Pcc2: per unit-bit-and-second of oRF
Ru: target rate of user u
β>=0: implies overhead power consumption of MU processing
compared to SU-MIMO
EE Maximization Problem
Splitting Problem
• Channel-gain-based greedy AS: RSSI
• Min-dist-based greedy AS: localization Info.
Cont.
• SINR-threshold-(γ)-based clustering
– SINR btw users in the same cluster < γ
– SINR btw users in diff clusters > γ
γ = 25dB
γ = 32dB
Per-Cluster Optimization
• Now, AS matrix is given
• For fixed PC matrix,
– ZF-MU-MIMO precoding matrix
Cont.
• Now, AS and precoding matrices are given
• Assumption: ICI is negligible
– Optimal for MU: using bisection algo, convex
feasibility test
– Heuristic for MU and optimal for SU:
Numerical Results
Single cell
Single antenna for each user
No adaptation for
- # of antennas for each user
- clustering threshold
Remaining Issues for L-DAS
• Deployment, implementation, and operation
–
–
–
–
–
–
Cell planning,
Regular/irregular deployment of antennas
Synchronization for large cluster
Robustness against CSI error
Infrastructure cost for wired optical fronthaul
Comparative, quantitative study of L-DAS and LCAS considering Capex and Opex
Cont.
• Iteration for
– # of antennas
– clustering threshold
Cont.
• Example at cell boundary of two cells
• Outage  increase # of active DAs
1
 =-Inf
0.9
 =-Inf
0
1
0.8
Circle: Non-outage user
Circle color stands for the cluster/DA
0.7
X: Deactivated DA
0.6
0.5
Colored Square:
Active distributed
antenna (DA)
0.4
0.3
Colored Thick Circle: Active DA
allocated to the outage user
0.2
Black Dot: outage user
Circle color stands for the cluster/DA
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Outage 
1
 =-Inf
0.9
 =-Inf
0
1
0.8
0.7
0.6
Threshold Update
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Increase clustering threshold γ  outage 
1
 =1e-005
0.9
 =1e-005
0
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Increase # of active DAs  outage 
1
 =1e-005
0.9
 =1e-005
0
1
0.8
0.7
0.6
Threshold Update
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Increase clustering threshold γ  outage 
1
 =4
0.9
 =4
0
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Increase # of active DAs  outage 
1
 =4
0.9
 =4
0
1
0.8
0.7
0.6
Threshold Update
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• No outage: threshold update (2,3) times
1
 =4
0.9
 =8
0
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cont.
• Demo
• cell_no_outage
Cont.
• Demo
• cell_outage