1. Quantized Precoding with Feedback 11n Partial Proposal
Month 2002
doc.: IEEE /xxxr0
September 2004
Quantized Precoding with Feedback 11n Partial Proposal
Robert W. Heath Jr., Johann Chiang,
Bishwarup Mondal and Roopsha Samanta
The University of Texas at Austin
Department of Electrical and Computer Engineering
Wireless Networking and Communications Group
1 University Station C0803, Austin, TX
Phone:
Fax:
{rheath, jchiang, mondal,
The University of Texas at Austin
John Doe, His Company

2. Outline Main features Closed-loop MIMO with limited feedback
September 2004
Outline
Main features
Closed-loop MIMO with limited feedback
Quantized precoding
Subcarrier clustering
Multi-mode precoding
MAC extension
Simulation results
Conclusion
The University of Texas at Austin

3. Main Features Closed-loop MIMO-OFDM with limited feedback (LF)
Month 2002
doc.: IEEE /xxxr0
September 2004
Main Features
Closed-loop MIMO-OFDM with limited feedback (LF)
Robust optional mode when RTS/CTS is on
Proposed PHY features
Quantized precoding with feedback
TX Beamforming (BF)
Spatial multiplexing (SM)
Multi-mode adaptation
Proposed MAC features
Extended RTS/CTS frames for feedback
Backward compatible with a
The University of Texas at Austin
John Doe, His Company

4. Proposal Overview Please refer to IEEE 802.11-04/962r0 September 2004
The University of Texas at Austin

5. Feedback Structure No physical feedback channels available in 802.11
Month 2002
doc.: IEEE /xxxr0
September 2004
Feedback Structure
No physical feedback channels available in
Exploit control frames in existing or emerging standards as logical feedback channels
Propose extension to existing MAC
Use RTS for estimation and CTS for feedback
The University of Texas at Austin
John Doe, His Company

6. When and Why Use RTS/CTS?
September 2004
When and Why Use RTS/CTS?
Efficient when # of active STAs is large
Reduce collision overhead in hotspots
Useful in DCF (no AP, peer-to-peer)
Alleviate hidden terminals issue
Low overhead when frame size is large
Benefit from frame aggregation
Required for backward compatibility
Should maximize the worth of legacy mechanism
Capable of Improving MAC throughput
Use dynamic RTS/CTS threshold (IEEE /312r0)
The University of Texas at Austin

7. MIMO without CSI at TX Space-time block code (STBC)
September 2004
MIMO without CSI at TX
Space-time block code (STBC)
Easily obtained spatial diversity
No array gain
Inflexible code design
Spatial multiplexing
Sensitive to channel invertibility
Limited spatial diversity
Hybrid mode (e.g. DSTTD)
Extra transmit antennas
The University of Texas at Austin

8. MIMO with CSI at TX Transmit beamforming Precoded spatial multiplexing
September 2004
MIMO with CSI at TX
Transmit beamforming
Significant spatial diversity
Additional array gain
Flexible implementation w/ various antenna configurations
Precoded spatial multiplexing
Improved channel invertibility w/ extra transmit antennas
Additional spatial diversity
Multi-mode precoding
Adaptation between precoded BF and SM modes
The University of Texas at Austin

9. How to Get CSI at TX? Open-loop: time division duplex (TDD)
September 2004
How to Get CSI at TX?
Open-loop: time division duplex (TDD)
Exploit channel reciprocity from control frames
Require RF calibration
Closed-loop: feedback
Inform transmitter about channel state
Require overhead for feedback
Can reduce overhead with efficient feedback
The University of Texas at Austin

10. Precoding H … … … Use more antennas than substreams
Month 2002
doc.: IEEE /xxxr0
September 2004
Precoding
Use more antennas than substreams
Apply channel dependent linear transform at the TX
Create effective channel with better invertibility
Use instantaneous channel state information (CSI)
Transmitter must be informed about H (or F) H
Why do we propose precoding? Performance of spatial multiplexing is very sensitive to channel singular values.
Precoding is a method of using additional antennas at the transmitter, beyond the number of substreams, to provide higher reliability and range through increased diversity. This advantage is achieved by customizing the transmitted signal to the channel; which requires the transmitter to be informed about the current channel state.
The question is how to reduce the size of feedback.. … …
Linear RX
Detection
Spatial
Multiplexing F …
Symbols
and
Decoding
Feedback Channel
Estimation &
Precoding
The University of Texas at Austin
John Doe, His Company

11. Month 2002
doc.: IEEE /xxxr0
September 2004
Quantized Precoding
Quantize precoder using instantaneous CSI (versus channel statistics)
Use fixed codebook of precoding matrices known to TX and RX
Select codeword from codebook and feedback index
Smart codebook designs based on Grassmannian subspace packing
The University of Texas at Austin
John Doe, His Company

12. Selection of Codebook Use Grassmannian precoding [Love & Heath]
September 2004
Selection of Codebook
Use Grassmannian precoding [Love & Heath]
Codebooks are optimal packings in the Grassmann manifold, G(Mt, M), where M is the number of data streams
Distance measure depends on precoding criteria
Set of subspaces
Codebooks available at
The University of Texas at Austin

13. Quantized Precoding with OFDM
Month 2002
doc.: IEEE /xxxr0
September 2004
Quantized Precoding with OFDM
P/S
&
Add CP
IDFT
Feedback of
quantized precoder
Remove
S/P
DFT
Vector
Decoder …
NOTE FOR JOAHNN: Explain the figure as a generic OFDM system using quantized precoding. Due to the number of tones, the feedback requirements increase. Hence we propose to quantize clusters of tones.
Based on the design issues discussed before, we can show that
Quantized precoding is a means of achieving the diversity gains of precoding with only a few bytes of feedback. It uses a predetermined codebook of (precoding) matrices known to both the transmitter and receiver. Based on an estimate of the channel state information, the receiver chooses a set of precoding matrices and sends the corresponding indices back to the transmitter. The transmitter reconstructs the precoding matrices from the indices and uses them for payload transmission. With a few bytes of feedback it is possible to realize near-ideal precoding gains with both beamforming and spatial multiplexing modes of operation.
Per tone model
The University of Texas at Austin
John Doe, His Company

14. Correlation of Subcarriers
Month 2002
doc.: IEEE /xxxr0
September 2004
Correlation of Subcarriers
Precoding in MIMO-OFDM requires
Feedback requirements  Number of subcarriers
How can we reduce the number of matrices fed back?
Exploit correlation between precoding matrices
Send back fraction of matrices and use smart interpolation
The University of Texas at Austin
John Doe, His Company

15. Clustering of Subcarriers
Month 2002
doc.: IEEE /xxxr0
September 2004
Clustering of Subcarriers
Subcarriers are grouped into clusters
Single codeword is chosen for each cluster
Clustering reduces feedback
The University of Texas at Austin
John Doe, His Company

16. Precoder Interpolation
Month 2002
doc.: IEEE /xxxr0
September 2004
Precoder Interpolation
Use codeword of the center subcarrier for cluster
Spherical interpolation with phase optimization
subcarriers
cluster 1
cluster 2
cluster 3
cluster 4
feedback
subcarriers
feedback …
Interpolate at the TX
The University of Texas at Austin
John Doe, His Company

17. System Diagram CTS Data RTS TX RX Feedback of codeword index
September 2004
System Diagram TX
Feedback of codeword index
Precoding at TX
Transmission of payload
When RTS/CTS is on
CTS
Data
RTS
Linear receiver and detection
Channel estimation at RX
RX selects precoder from codebook RX
The University of Texas at Austin

18. Multi-Mode Precoding Allow variable number of substreams
September 2004
Multi-Mode Precoding
Allow variable number of substreams
Select optimal # of substreams and rate jointly
Use a heterogeneous codebook
Provide flexible diversity-multiplexing tradeoff
The University of Texas at Austin

19. Multi-Mode Implementation
Month 2002
doc.: IEEE /xxxr0
September 2004
Multi-Mode Implementation
Number of STA antennas
Mandatory mode (2 antennas)
Optional modes (3 or 4 antennas)
Mode / number of substreams
Precoded beamforming (1 substream)
Precoded spatial multiplexing (>1 substreams)
Multi-mode MAC adaptation (frame-based)
Feedback the mode selection bit field
Multi-mode PHY adaptation (cluster-based)
Embed information of number of substreams in codebook
The University of Texas at Austin
John Doe, His Company

20. Multi-Mode Clustering
September 2004
Multi-Mode Clustering
Enable fast PHY adaptation
Adaptive loading (Modulation/Coding/Substream)
Substream adaptation based on channel condition
Rate adaptation based on SNR
Embed both adaptation mode information in codebook
The University of Texas at Austin

21. MAC Extension Extension to legacy RTS frame
September 2004
MAC Extension
Extension to legacy RTS frame
Append training sequences for multiple antennas
Extension to legacy CTS frame
Exploit free 10 bit available, required to fill OFDM symbol
Mode selection
Use higher order modulation (QPSK) for limited feedback
48 bits => 1 OFDM symbol
The University of Texas at Austin

22. Bit Field Format CTS Feedback format Mode selection field format
September 2004
Bit Field Format
CTS Feedback format
Mode selection field format
Mode
Antenna (APxSTA)
# Substreams
Data Rate (Mbps)
000
4x2 1
6/9/12/18/24/36/48/54
001 2
12/18/24/48/72/96/108
010
4x3
011
100 3
18/27/36/54/72/108/144/162
101
4x4
110
111
The University of Texas at Austin

23. Spectral Efficiency of SM
September 2004
Spectral Efficiency of SM
4x2/4x3/4x4 precoded SM with 2 substreams, 64QAM R3/4 (108Mbps)
Frame Size
1000B
2000B
3000B
4000B
Payload
5.2632
5.3571
5.3333
+Preamble
3.5714
4.2553
4.6154
4.7619
+ACK
2.6991
3.5682
4.0513
4.2988
+RTS/CTS
1.8128
2.6968
3.2551
3.5983
+Feedback
1.7193
2.5919
3.1525
3.5038
Overhead
36.30%
27.36%
22.19%
18.49%
The University of Texas at Austin

24. Spectral Efficiency of BF
September 2004
Spectral Efficiency of BF
4x2/4x3/4x4 TX BF w/ LF, 64QAM R3/4 (54Mbps)
Frame Size
1000B
2000B
3000B
4000B
Payload
2.6316
2.6667
2.6786
2.6846
+Preamble
2.1277
2.3810
2.4793
2.5316
+ACK
1.7841
2.1494
2.3068
2.3945
+RTS/CTS
1.3484
1.7992
2.0248
2.1603
+Feedback
1.2960
1.7519
1.9846
2.1258
Overhead
27.36%
18.49%
13.97%
11.22%
The University of Texas at Austin

25. Comparison Assumption
September 2004
Comparison Assumption
Assume no collision
All retransmission results from packet error
In fact, RTS/CTS can reduce collision overhead
Open-loop
Retransmission overhead depending on PER
Closed-loop
Fixed overhead from RTS/CTS and feedback
Equivalent points for both overhead
SM: PER=0.3630
BF: PER=0.2736
The University of Texas at Austin

26. Simulation Setup Substream number: 2 (SM)
September 2004
Simulation Setup
Substream number: 2 (SM)
Precoding quantization: 6 bits
Subcarrier clustering size: 6
Feedback amount: 48 bit
Modulation: 64QAM
Coding rate: 3/4
Data Rate: 108 Mbps
Frame size: 1000 Byte
Channel Model: Ch B, Ch D, Ch E
The University of Texas at Austin

27. SM with 2 Substreams in Ch B
September 2004
SM with 2 Substreams in Ch B
5 dB
The University of Texas at Austin

28. SM with 2 Substreams in Ch D
September 2004
SM with 2 Substreams in Ch D
5 dB
The University of Texas at Austin

29. SM with 2 Substreams in Ch E
September 2004
SM with 2 Substreams in Ch E
6 dB
The University of Texas at Austin

30. September 2004
Conclusions
Quantized precoding with feedback is practical to improve goodput even if RTS/CTS is on
Subcarrier clustering reduces feedback for OFDM
Multi-mode precoding provides for diversity- multiplexing tradeoff and compatible with rate adaptation
Q&A
The University of Texas at Austin

31. References Webpage: http://www.ece.utexas.edu/~rheath/lf/
September 2004
References
Webpage:
D. J. Love, R. W. Heath, Jr., W. Santipach, and M. L. Honig, "What is the Value of Limited Feedback for MIMO Channels?," to appear in IEEE Communications Magazine
J. Choi and R. W. Heath, Jr., "Interpolation Based Transmit Beamforming for MIMO-OFDM with Limited Feedback,'' Proc. of IEEE Int. Conf. on Communications, Paris, France, June 2004.
R. W. Heath Jr., "Precoding and Interpolation for Spatial Multiplexing MIMO-OFDM with Limited Feedback,'' The 2004 Workshop on Smart Antennas in Communications, Stanford, California, July
The University of Texas at Austin

32. September 2004
References
D. J. Love, R. W. Heath Jr., and T. Strohmer, “Grassmannian Beamforming for Multiple-Input Multiple-Output Wireless Systems,” IEEE Trans. Inf. Th., vol. 49, pp , Oct
D. J. Love and R. W. Heath Jr., “Necessary and Sufficient Conditions for Full Diversity Order in Correlated Rayleigh Fading Beamforming and Combining Systems,” to appear in IEEE Trans. Wireless Comm.
D. J. Love and R. W. Heath Jr., “Limited feedback unitary precoding for spatial multiplexing systems,” submitted to IEEE Trans. Inf. Th., July 2003 (see also Globecom 2003)
R. W. Heath Jr. and D. J. Love, “Multi-mode precoding for spatial multiplexing systems,” submitted to IEEE Trans. Sig. Proc., December 2003 (see also Allerton 2003)
The University of Texas at Austin