Transcript IEEE C802.16m-08/379

Multiple user MIMO with only precoded pilot and limited feedback for FDD DL
Document Number: C802.16m-08/379
Date Submitted: 2008-05-05
Source:
Dongjun Lee, Jungnam Yun, RajagopalS.Iyengar
POSDATA Co, Ltd.
276-2, Seohyeon-dong, Bundang-gu, Seongnam-si,
Kyeonggi-do,463-775, Korea
Jungwoo Lee,
School of Electrical Engineering,
Seoul National University
Korea
Email: [email protected]
Email: [email protected]
Venue:
IEEE 802.16m-08/005 “Call for Contributions on Project 802.16m System Description Document (SDD)”, In response to the following
topics: “DL MIMO”
Purpose:
To be discussed and adopted by TGm for the 802.16m SDD.
Notice:
This document does not represent the agreed views of the IEEE 802.16 Working Group or any of its subgroups. It represents only the views of the participants listed in
the “Source(s)” field above. It is offered as a basis for discussion. It is not binding on the contributor(s), who reserve(s) the right to add, amend or withdraw material
contained herein.
Release:
The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an
IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s
sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this
contribution may be made public by IEEE 802.16.
Patent Policy:
The contributor is familiar with the IEEE-SA Patent Policy and Procedures:
<http://standards.ieee.org/guides/bylaws/sect6-7.html#6> and <http://standards.ieee.org/guides/opman/sect6.html#6.3>.
Further information is located at <http://standards.ieee.org/board/pat/pat-material.html> and <http://standards.ieee.org/board/pat >.
1
Overview
 This contribution presents a new MU-MIMO scheme for FDD DL
 In the conventional MU-MIMO for FDD DL, there are disadvantages in 2
aspects.
1. Channel Feedback need too many bits for better performance. Reduced
feedback rate leads to great performance loss.
2. Both common pilot and precoded dedicated pilot are needed. Common pilot is
used for channel feedback, while precoded dedicated pilot is for MIMO
detection. There will be too much DL over head.
 We propose a novel MU-MIMO scheme for FDD DL, where only
precoded pilot is needed for transmission side. To reduce channel
feedback, we combine the channel element scalar quantization, differential
channel feedback, and Rx antenna selection which can greatly reduce the
feedback bits.
Feedback Strategy for Multiuser MIMO
• Codebook approach
– Small feedback bits (Codebook)
– High complexity
• Quantization (scalar) approach
– Many feedback bits (Quantization per element)
– Cartesian coordinate is desirable.
– Low complexity
• Scalar quantization is preferred.
– Due to the need for precise CSI for MU-MIMO to
remove inter-user interferences
Channel Element Based Scalar quantization
• Scalar quantization per channel element
– Rectangular coordinates
– Polar coordinates:
total 8 bits
• Magnitude bits : 3 bits
• Phase bits : 5 bits
– Rectangular coordinate Is
preferred.
M = 4, N = 2, Kt = 30
35
30
Sum Capacity (bits/s/Hz)
• 8 bits
• 4 bits for real part
• 4bits for image part
rectangular B = 4
polar MB = 2 PB = 6
polar MB = 1 PB = 7
polar MB = 3 PB = 5
25
20
15
10
5
0
0
5
10
15
SNR
20
25
30
Performance of Channel Element Based Scalar
quantization
• Scalar quantization per each element
– 8 bits per each element (4 bits for real part, 4bits for image)
– 32 bits for 4ⅹ1 channel matrix
M = 4, N = 2, Kt = 30, B = 3, BER, Coordinated MMSE-VBLAST
10
0
M = 4, N = 2, Kt = 30, B = 4, BER, Coordinated, MMSE-VBLAST, SVD
0
10
multi users selection coordinated
1 user selection
multi users selection
1 user selection
-1
10
10
-1
-2
10
BER
BER
10
-2
10
-3
10
-3
-4
10
10
-4
-5
10
10
-5
0
-6
5
10
15
SNR
20
25
30
10
0
2
4
6
8
10
SNR
12
14
16
18
20
Feedback Reduction Schemes
Pros
• Usage of Differential Channel Feedback to reduce
UL overhead
 Only Feed back the whole channel matrix at the initial
transmission
 In the later transmission, MS only feedbacks 1-2 bits
differential information for each channel element.
• Rx antenna selection
 Reduce channel feedback by effective converting a 4x2
MIMO channel into a 4x1 MIMO channel.
 Reduce feedback rate by factor of 2.
Feedback Reduction for Multiple
Rx Antennas
• Pros
– Reduce the feedback rate by factor of 2 by using channel matrix size
from 4x2 to 4x1
– Increase simultaneously serviced users from 2 to 4: increased multiuser diversity
• Cons:
– peak DL rate for individual user is halved.
Feedback Reduction for Multiple
Rx Antennas
• Antenna selection mode
– Select one of many Rx antennas
M = 4, N = 2, Kt = 30, perfect CSI
M = 4, N = 2, Kt = 30, scalar quantization
45
40
35
stream per MS = 2
Antenna selection
stream per MS = 2
Antenna selection
30
Sum Capacity (bits/s/Hz)
Sum Capacity (bits/s/Hz)
35
30
25
20
15
25
20
15
10
10
5
5
0
0
5
10
15
SNR
20
25
30
0
0
5
10
15
SNR
20
25
30
MU-MIMO with Only precoded pilot
• Precoded pilots: two approaches
– Accumulated precoding
– Precoding inversion
Advantage of MU-MIMO with Only precoded pilot
Pros
• Reduce DL overhead due to;
 Only Precoded common pilots are needed for both
channel feedback and detection
 No need for both precoded and non-precoded pilot
Cons
• Performance loss by multi user scheduling based
on the precoded pilot this can be further
optimized by accumulated precoding and
precoding matrix inverse
Accumulated Precoding for MU-MIMO
Acumulate
Precoding
Matrix
MS 1
MS 1
Data
MS 2
Data
Scheduling
MS K
Data
Precoding
Matrix
Acumulate
Precoding
Matrix
of
Previous
Frame
MS 2
MS K
Precoded CSI
Feedback
Figure 2. Accumulated precoding for MU-MIMO
-
In FDD system, following operations can be applied as in Figure 2.
-
the BS only transmits the precoded dedicate pilot. By this precoded dedicate pilot,
MS estimates his precoded channel information and feedback to the BS.
Based on all the MS’s precoded CSI information, BS does the scheduling and
calculates the corresponding new precoding matrix for the selected MSs.
The final precoding matrix is the combination of the calculated new precoding matrix
with the previous frame’s precoding matrix—Acumulated precoding.
Accumulated Precoding
• Accumulated precoding
 H1V11
 0

 0

H  0
 H 5V11


H V
 KT 11
0
H 2V21
0
0
H 5V21
0
0
H 3V31
0
H 5V31
H KT V21
H KT V31
  H1V1 
 HV 
  2 1
  H 3V1 
 

H
V

  4 1
  H 5V1 
 

 

H KT V41   H KT V1 
0
0
0
H 4V41
H 5V41
where Vik : precoding matrix for user i at frame k
– V j  [V1 j V2 j V3 j V4 j ]
Accumulated Precoding
M = 4, N = 2, Kt = 30, Perfect CSI, 1 frame per coherence time
M = 4, N = 2, Kt = 30, Perfect CSI, 3 frame per coherence time
45
40
45
antenna selection
accumulated precoding
40
35
Sum Capacity (bits/s/Hz)
Sum Capacity (bits/s/Hz)
35
30
25
20
15
10
30
25
20
15
10
5
0
0
antenna selection
accumulated precoding
5
5
10
15
SNR
20
25
30
0
0
5
10
15
SNR
20
25
30
Accumulated Precoding
• Performance degrades when coherence time gets
longer
– Not good for nomadic or low-mobility users
M = 4, N = 2, Kt = 30, Perfect CSI, 10 frame per coherence time
M = 4, N = 2, Kt = 30, Perfect CSI, 5 frame per coherence time
45
40
45
antenna selection
accumulated precoding
40
35
Sum Capacity (bits/s/Hz)
Sum Capacity (bits/s/Hz)
35
30
25
20
15
30
25
20
15
10
10
5
5
0
0
antenna selection
accumulated precoding
5
10
15
SNR
20
25
30
0
0
5
10
15
SNR
20
25
30
Power-adjusted Accumulated Precoding
• A problem with accumulated precoding
– In the scheduling for next frame with H , previously
selected users is not favored because of 0 in channel
element, which reduces the effective channel power
• Solution: power-adjusted accumulated precoding
– Multiply H by 1/2 for non-selected users for channels
with coherence time of larger than 1 frame.
– Multiplication by ½ needs to be done adaptively
depending on the coherence time.
– Difficult to implement
Power-adjusted Accumulated Precoding
• Power-adjusted accumulated precoding
0
0
0
 H1V11
  H1V1 

  HV 
0
H
V
0
0
2
21
2 1

 


  H 3V1 
0
0
H 3V31
0
'

 

0
0
0
H 4V41    H 4V1 
H 
 0.5H 5V11 0.5H 5V21 0.5H 5V31 0.5H 5V41   0.5H 5V1 

 


 

0.5H V 0.5H V
 0.5H V 
0.5
H
V
0.5
H
V
K
11
K
21
K
31
K
41
KT 1 

T
T
T
T
 
– Select users with H
'
Power-adjusted Accumulated Precoding
• Simulation results
– It works well but needs to estimate the coherence time, and
adapt the algorithm.
M = 4, N = 2, Kt = 30, Perfect CSI, 5 frame per coherence time
M = 4, N = 2, Kt = 30, Perfect CSI, 10 frame per coherence time
45
40
45
antenna selection
accumulated precoding
40
35
Sum Capacity (bits/s/Hz)
Sum Capacity (bits/s/Hz)
35
30
25
20
15
30
25
20
15
10
10
5
5
0
0
antenna selection
accumulated precoding
5
10
15
SNR
20
25
30
0
0
5
10
15
SNR
20
25
30
Precoding Inversion Scheme for MU-MIMO
• Channel estimation at ith user
– H i (V11  V22  V33  V44 )
1 , 2 , 3 , 4 are orthonomal pilot each other
– H (V   V   V   V  )  [ '  '  '  ']
i
1 1
2 2
3 3
4 4
1
2
3
4
 [ H iV1 H iV2 H iV3 H iV4 ] --- (A)
– (A) is fed back to the BS
Precoding Inversion Scheme
• At the BS
– [ H V H V H V H V ]  H  [V V V V ]  H V
i 1
i 2
i 3
i 4
i
1 2
3
4
i
where [V1 V2 V3 V4 ]  V
–
H i  [ H iV1 H iV2 H iV3 H iV4 ]V 1
• We empirically observe that V 1 exists through
simulations
• Therefore we can estimate the raw channel through
precoding matrix inversion at the BS
Precoding Inversion Scheme
• Simulation results
– No need to adapt the algorithm based on coherence time
• Precoding inversion is preferred over the accumulated
approach due to the adaptation requirement
M = 4, N = 2, Kt = 30, Perfect CSI, 1 frame per coherence time
M = 4, N = 2, Kt = 30, Perfect CSI, 10 frame per coherence time
45
40
45
antenna selection
inverse precoding
40
35
Sum Capacity (bits/s/Hz)
Sum Capacity (bits/s/Hz)
35
30
25
20
15
30
25
20
15
10
10
5
5
0
0
antenna selection
inverse precoding
5
10
15
SNR
20
25
30
0
0
5
10
15
SNR
20
25
30
Summary of Proposals
• Quantization is preferred over codebook due to lower search
complexity
– 8 bits per complex channel matrix elements using
rectangular coordinate
– # of feedback bits: 32 bits per subcarrier group per frame
per user
– differential channel feedback can be used to further reduce
the overall amount of feedback
• Precoded pilots
– Precoding inversion is preferred over the accumulated
precoding because no adaptation is required for precoding
inversion.
Text Recommendations for SDD
Section 11 Physical Layer
• Section 11.x DL MIMO structure
– 11.x.y DL MU-MIMO structure for FDD
• DL MU-MIMO with precoded pilot only shall be used for FDD mode to
reduce DL overhead
• Channel element scalar quantization, differential channel feedback and
antenna selection are used so that the UL overhead can be reduced
22