Beamforming training for IEEE 802.11ad Date

May 2010
doc.: IEEE 802.11-10/0493r1
Beamforming training for IEEE 802.11ad
Date: 2010-05-17
Authors:
Name
Company
Address
Phone
Changsoon Choi
Eckhard Grass
email
+493355625155
IHP
Im Technologiepark 25,
Frankfurt (oder), Germany
Rolf Kraemer
Thomas Derham
Sandrine Roblot
Orange Labs
Orange Labs Tokyo,
Shinjuku 160-0022
4, rue du clos courtel,
35512 Cesson-Sevigne
Laurent Cariou
Philippe Christin
Submission
Slide 1
+81-3-5312-8563
[email protected]
[email protected]
[email protected]
thomas.derham@
orange-ftgroup.com
sandrine.roblot@
orange-ftgroup.com
laurent.cariou@
orange-ftgroup.com
philippe.christin@
orange-ftgroup.com
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Abstract
• The performance of 60-GHz wireless LAN can be significantly
enhanced if the receiver beamforming is capable of interference
mitigation.
• In order to do this, beamforming training mechanism should allow
for estimation of the CSI (channel state information) matrix.
• This proposal addresses the number of beamforming training
sequence repetition necessary to achieve this, and demonstrates
the performance improvement that can be obtained.
Submission
Slide 2
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Beamforming for interference mitigation
•
•
•
•
Important to manage mutual interference among different 60-GHz devices /networks.
Even within TGad networks, interference is a main concern for efficient spatial reuse.
Beamforming (BF) needs interference mitigation capability.
IEEE 802.15.3c BF is NOT capable of it due to the nature of codebook approach
•
In order to achieve interference mitigation, there should be a mechanism in 802.11ad for
the channel matrix to be estimated
e.g. IEEE 802.15.3c
AP
Interference
STA
Submission
Slide 3
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Beamforming for < 6-GHz and 60-GHz
60-GHz
< 6-GHz
Digital
baseband
Digital
baseband
Digital
baseband
Digital
baseband
Digital
baseband
Analog phase-shifter
Weighting
vector
calculation
•
•
60-GHz BF transceivers would be based on analog beamforming
Baseband does not know the received signals on each antenna individually because
they are combined in analog domain prior to digital baseband
 elements of MIMO channel matrix cannot be estimated directly
Submission
Slide 4
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
BF training proposal
• For BF training of an N-element receiver STA, a transmit STA will
send N-repetitions of BF training sequences for one Tx beam.
• Receiver STA can estimate channel state information (CSI) in various
ways (e.g. LS, MMSE).
For N-element beamforming receiver
BF training time
BF training
symbol #2
SBIFS
SBIFS
BF training
symbol #1
BF training
symbol #N
time
Submission
Slide 5
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
BF model for performance evaluation
• Consider SIMO channel.
• This reflects the usage case where one mobile terminal (e.g. smart
phone) transmits data to an access point with beamforming
capability.
Tx
Rx
x
c1*
h1
h2
Non- beamforming capable
Beamforming capable
c2*
h3
c3*
y
Digital
baseband
hN
c*N
Submission
Slide 6
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Example: BF training with codebook
approach
• Transmit STA sends N-repetitions of a BF training sequence while
the receiver cycles through different beamforming vectors from
codebook matrix (C)
– Codebook matrix (n-element, k-beam) defined as:
C  [c1 , c 2 ,c k ]
c  [c1 , cn ]T
C  (n  k )
matrix
• Received baseband signals for k-th beamforming vector
 
y c
k
k H
H  x  nk
H  [h1 ,h n ]T
• Collect all baseband signals (or channel estimates) for n-repetition
BF training sequences
Y  CH H  X  N
Y, C  ( n  k )  ( n  n )
• Estimation of CSI on each antenna
Submission
Slide 7
matrix, n = k for matrix inversion
Ĥ  [C H ]1 Y  X*
Changsoon Choi, IHP microelectronics
doc.: IEEE 802.11-10/0493r1
System simulation model for BF evaluation
0
0
-10
-10
-20
-30
-40
-20
-30
-40
-50
-50
-60
-60
0
20
40
60
time index
80
-100
0
100
angle-of-arrival [deg]
Antenna (90-degree HPBW)
90
Beam pattern for codebook (C)
4 90
1
120
60
60
0.8
120
3
0.6
150
30
2
30
150
0.4
1
0.2
180
0
210
330
240
300
270
Submission
Angular response
10
relative response [dB]
Channel and antenna models
• 60-GHz NLOS residential model
(CM2.3) with AoA information (used in
IEEE 802.15.3c)
• 100 channel realizations and averaged
results. Each channel normalized to unit
power
• 90-degree Gaussian beam pattern
HPBW (half-power beamwidth) for
receiver antenna. No backside emission
assumed.
• Constant total gain from beamformers
assumed
• BF codebook matrix (C) from IEEE
802.15.3c std
relative response [dB]
Time response
10
180
0
330
210
300
240
270
Changsoon Choi, IHP microelectronics
doc.: IEEE 802.11-10/0493r1
BF performance with full CSI
(no interference)
Beamforming gain vs. number of RX antennas
10
•
Maximum signal-to-interference plus
noise (SINR) beamformer is used for
this work.
Codebook: IEEE 802.15.3c standard
Beamforming gain [dB]
9
8
2
No interference
SNR = 10-dB
SINR 
7
6

5
E[ w H hs ]
2
2
E[ w H i ]  E [ w H n ]

E[w H hss H h H w ]
E[w H ii H w ]  E[w H nn H w ]
CSI covariance matrix
 w E[hh ] w
 w R hh w
 H
H
2
H
w E[ii ]w   n w w w R ii w   n2 w H w
2
s
H
H H
2
s
H
H
H
Interference covariance
4
This work, AWGN
This work, 60-GHz CM2.3
Codebook, AWGN
Codebook, 60-GHz CM2.3
3
2
1
2
4
6
8
Number of antenna element
Submission
10
•
•
IEEE 802.15.3c beamformer is included
for comparison
Improved beamforming gain is
obtained with full MIMO CSI
Changsoon Choi, IHP microelectronics
doc.: IEEE 802.11-10/0493r1
BF performance with full CSI
(with co-channel interference)
10
This work
Codebook
CIR
relative response [dB]
5
Interference
at 45-degree
0
-5
-10
-15
-20
-100
-80
-60
-40
-20
0
20
angle-of-arrival [deg]
•
60
80
100
40
Codebook: IEEE 802.15.3c standard
35
30
Interference at 45-deg AoA
Input SIR = 6-dB
25
20
15
10
This work, AWGN
This work, 60-GHz CM2.3
5
Codebook, AWGN
Codebook, 60-GHz CM2.3
0
0
5
10
15
20
25
30
35
Input SNR per element [dB]
Co-channel interference
–
–
•
•
•
40
Output SINR vs. Input SNR
Output signal-to-interference noise ratio (SINR) [dB]
Array factors for full CSI beamforming and codebook
Assume that angle of arrival (AoA) of co-channel interference was ideally estimated in receiver
Random signals (AWGN-like) with random AoA were generated for co-channel interference.
Beamforming provides efficient interference nulling with full MIMO CSI.
Higher SINR can be expected with the help of interference mitigation.
No interference mitigation capability in IEEE 802.15.3c codebook BF.
Submission
Changsoon Choi, IHP microelectronics
doc.: IEEE 802.11-10/0493r1
Optimization of Tx and Rx beamforming
vectors (1) – SIMO and MISO channels
•
Method for estimating SIMO channel can be used for MISO channel.
– Tx has M elements, Rx has N elements
•
•
Find best beams (BF vectors) for Tx and Rx by switching different beams
For fixed Tx BF vectors
– Tx transmits N repetitions of training sequence
• For each repetition, receive STA uses a different beamforming vector from codebook
– Optimize Rx BF vectors using the estimated SIMO channel matrix
• Optimization algorithm (e.g. Max SINR, MMSE) is implementation-dependent
•
For optimized Rx BF vectors (through above-mentioned process)
– Tx transmits M repetitions of training sequence
• For each repetition, transmit STA uses a different beamforming vector from codebook
– Estimated CSIs for different Tx BF vectors are fed back to Tx
– Optimized Tx BF vectors using the estimated SIMO channel matrix
•
•
M  N repetition of training sequences
This procedure can be repeated in multiple times for maximizing SINR
Submission
May 2010
doc.: IEEE 802.11-10/0493r1
Optimization of Tx and Rx beamforming
vectors (2) – Full MIMO channels
•
•
Method for estimating SIMO channel can be extended to MIMO
Transmit STA sends:
–
N
repetitions of training sequence
• for each repetition, receive STA uses a different beamforming vector from codebook
matrix
– where the above repetitions are repeated
M
times
• for each repetition, transmit STA uses a different beamforming vector from codebook
matrix
– Codebook matrices should be orthogonal
•
•
•
•
Complex received signal on subcarrier i for each repetition placed in
corresponding element of matrix
Full MIMO channel state information Yi
H 1
1
*
M  N repetition of BF training is required. Ĥi  [C ] Yi W  X
Maximum performance but higher complexity
Submission
Slide 12
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Conclusion
• This proposal addresses required number of
beamforming training sequences for channel matrix
estimation (SIMO, MISO).
• It gives us possibility to adaptively mitigate co-channel
interference, which is also advantageous for spatial
reuse.
Submission
Slide 13
Changsoon Choi, IHP microelectronics
May 2010
doc.: IEEE 802.11-10/0493r1
Acknowledgement
• This work has been supported by the European Community’s
Seventh Framework Programs referred to as MIMAX and
OMEGA
Submission
Slide 14
Changsoon Choi, IHP microelectronics