1 Analog Feedback for IEEE 802.16m - Performance Evaluation and Analog Feedback for IEEE 802.16m - Performance Evaluation and Design Details Design Details Document Number: IEEE S802.16m-09/0912 Date Submitted: 2009-04-27 Source: Fred Vook, Eugene Visotsky, Bill Hillery, Tim Thomas, E-mail: [email protected]Mark Cudak, Bishwarup Mondal, Fan Wang Motorola Inc. *<http:// standards.ieee.org/faqs/affiliationFAQ.html > Venue: IEEE 802.16m AWD – Call for Contributions IEEE 802.16m-09/0020 Base Contribution: C802.16m-09/0912 Abstract: This presentation discusses the design details and performance evaluation of analog feedback for IEEE 802.16m. Various design issues are also discussed. The proposed amendment text is contained in the base contribution. Purpose: Discuss and adopt proposed text contained in base contribution 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 >.
Transcript
1
Analog Feedback for IEEE 802.16m - Performance Evaluation and Design DetailsAnalog Feedback for IEEE 802.16m - Performance Evaluation and Design Details
Document Number:IEEE S802.16m-09/0912
Date Submitted:2009-04-27
Source: Fred Vook, Eugene Visotsky, Bill Hillery, Tim Thomas, E-mail: [email protected] Mark Cudak, Bishwarup Mondal, Fan Wang
Motorola Inc. *<http://standards.ieee.org/faqs/affiliationFAQ.html>
Venue:IEEE 802.16m AWD – Call for Contributions IEEE 802.16m-09/0020
Base Contribution: C802.16m-09/0912Abstract:
This presentation discusses the design details and performance evaluation of analog feedback for IEEE 802.16m. Various design issues are also discussed. The proposed amendment text is contained in the base contribution.
Purpose:Discuss and adopt proposed text contained in base contribution
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 >.
2
Overview
• Detailed design for analog feedback channels
• Overhead comparison between analog feedback and base & adaptive codebooks
• System level performance comparison between analog feedback and base & adaptive codebooks
3
Analog Feedback Channels within UL Control• Define an Analog Feedback (AFB) channel to be another type of UL
feedback channel– Some number of distributed LRUs are assigned to be AFB channel
• UL Control channels are currently assigned to distributed LRUs– A Distributed LRU is three 6x6 tiles
• The distributed LRUs on UL are then to be divided into:– Data– BW request channel
• Three 6x6 tiles equals a BW request channel– Feedback channels
• HARQ-ACK/NAK, • Primary & Secondary Fast Feedback• Distributed LRUs assigned to feedback channels are divided into 2x6 Feedback Mini
Tiles (FMTs)• Three permuted 2x6 FMTs equals a feedback channel
– Analog feedback channels (Our proposal)• Each distributed LRU that is assigned to be an analog feedback channel forms an
“analog feedback channel”.• Three 6x6 tiles equals an analog feedback channel
4
Analog Feedback Channel
• The BS assigns some number of distributed LRUs to be Analog Feedback LRUs (AFB LRUs)– MZONE: Each analog feedback LRU is comprised of three 6x6 tiles– LZONE: Each analog feedback LRU is comprised of three 4x6 tiles
• One AFB LRU contains Nafb analog feedback subchannels, where Nafb depends on the number of BS TX antennas:– Nafb=6 for BS=2TX– Nafb=6 for BS=4TX– Nafb=4 for BS=8TX
• Each AFB subchannel is designed to carry eigenvector feedback for one user– The intent is to support MU-MIMO or rank-1 SU-MIMO
• AFB subchannels are multiplexed via combination of FDM / TDM / CDM– CDM spreading / de-spreading is employed to improve post-combining
SINR
5
Data to be mapped onto an AFB subchannel
• MS estimates DL TX Covariance Matrix via DL multi-antenna pilots (e.g., MIMO midamble, DL reference pilots) over the entire band or over a sub-band
• MS is told whether to calculate a wideband eigenvector or a narrowband eigenvector– Wideband eigenvector is the dominant eigenvector of the
covariance matrix computed over the entire DL bandwidth• Supports rank1 BF / MU-MIMO over mini-band CRUs
– Narrowband eigenvector is the dominant eigenvector of the covariance matrix computed over a sub-band
• BS indicates which sub-band the MS must measure in the command to send the analog feedback
• Supports rank 1 BF / MU-MIMO over sub-band CRUs
• Analog symbols to be sent back (N BS TX antennas)– Eigenvector entries: e1 e2 … eN
– Eigenvector entries are then CDM spread and mapped to the assigned AFB subchannel (next…)
6
CDM Spreading of Analog Feedback with Mutually Unbiased Bases (MUBs)
• Performance of Analog feedback-based methods is sensitive to the UL CINR
• Performance can be improved by CDM spreading the analog feedback
• Use spreading codes designed from mutually unbiased bases (MUB)– Any two codes within a particular MUB set are orthogonal– Any two codes belonging to two different MUB sets (e.g., A, B) are
guaranteed to have a low cross-correlation level which results in a low interference level
• A sector is assigned a particular MUB matrix– Indicated via TBD
• Values to be mapped (see next slides): eijk = mji*ek, – ith MUB sequence within the assigned MUB matrix: mji, j=1:D– e1 e2 … eN are the eigenvector values for N BS TX antennas
BbAaba any and ,any for ,/1,2
D
7
Example MUBs (D=2,4)
• D=2
• D=4
,11
2
1,
11
11
2
1
iiBA
iiii
iiii
iiii
iiii
iiii
iiii
1111
1111
2
1,
1111
1111
2
1
,1111
1111
2
1,
1111
1111
1111
1111
2
1
DC
BA
8
Example MUB (D=8)
• D=8
11111111
11111111
11111111
11111111
11111111
11111111
11111111
11111111
8
1A AEADACAB
i
i
i
idiag
i
i
i
i
diag
i
i
i
i
diag
i
i
i
i
diag1
1
1
1
,
1
1
1
1
,
1
1
1
1
,
1
1
1
1
AHAGAF
1
1
1
1
,
1
1
1
1
,
1
1
1
1
i
i
i
i
diag
i
i
i
i
diag
i
i
i
i
diag
9
Analog feedback subchannel definition for M-Zone:BS=2 TX: Nafb=6, CDM factor D=4
2 Tx antennas• Nafb=6 analog feedback subchannels, each having a CDM factor of 4
• Mapping for 1st 6x6 tile is shown
• Mapping on 2nd and 3rd 6x6 tile is identical to 1st tile
• en is the nth component of the eigenvector for the given user
• Pk is a pilot symbol
• CDM spreading:
•Having e1 switch places with e2 on alternating frequency groups keeps average transmit power on the OFDM symbols the same
•Blank squares means the user sends nothing on those subcarriers/OFDM symbols
• Overall maximum gain is 48 (2 receive antennas times spreading of 4 times repetition of 6 across frequency)
4 Tx antennas• Nafb=4 analog feedback subchannels, each having a CDM factor of 4
• Mapping for 1st and 2nd 4x6 tile is shown. Mapping on 3rd 4x6 tile is identical to 1st tile
• en is the nth component of the eigenvector for the given user
• Pk is a pilot symbol
• CDM spreading:
• Blank squares means the MS sends nothing on those subcarriers/OFDM symbols
• Overall maximum gain is 48 (4 receive antennas times spreading of 4 times repetition of 3 across frequency)
eijk = mji*ek pijk = mji*pk
14
Analog feedback subchannel definition for L-Zone:BS=8 TX: Nafb=3, CDM factor D=2
AMS 1
AF
B T
ile 1
AF
B T
ile 2
AF
B T
ile 3
AMS 2 AMS 3
e111 e121 e112 e122p111 p121
e113 e123 e114 e124p112 p122
e115 e125 e116 e126p113 p123
p114 p124e117 e127 e118 e128
e112 e122
e114 e124
e116 e126
e118 e128
e111 e121
e113 e123
e115 e125
e117 e127
p111 p121
p112 p122
p113 p123
p114 p124
e211 e221 e212 e222p211 p221
e213 e223 e214 e224p212 p222
e215 e225 e216 e226
e217 e227 e218 e228
p213 p223
e112 e122
e114 e124
e116 e126
e118 e128
e111 e121
e113 e123
e115 e125
e117 e127
p111 p121
p112 p122
p113 p123
p114 p124
p214 p224
e211 e221 e212 e222p211 p221
e213 e223 e214 e224p212 p222
e215 e225 e216 e226
e217 e227 e218 e228
p213 p223
p214 p224
p211 p221
p212 p222
p213 p223
p214 p224
e212 e222
e214 e224
e216 e226
e218 e228
e211 e221
e213 e223
e215 e225
e217 e227
8 Tx antennas• Nafb=3 analog feedback subchannels, each having a CDM factor of 4
• en is the nth component of the eigenvector for the given user
• Having e1 (e3,e5,e7) switch places with e2 (e4,e6,e8) on alternating frequency groups keeps average transmit power on the OFDM symbols the same
• Pk is a pilot symbol
•Blank squares means the user sends nothing on those subcarriers/OFDM symbols
•Overall maximum gain is 64 (8 receive antennas times spreading of 4 times repetition of 2 across frequency)
eijk = mji*ek
pijk = mji*pk
15
UL Transmission Considerations
• Sending all N entries of the eigenvector (rather than normalizing to one of the entries) enables the eigenvectors to be power-scaled independently of the pilots
• The total TX power is determined by power control strategy
• Must apply scale factor across the band to insure total TX power is maintained
• Must preserve gain relationships between the entries of the eigenvector
16
Overhead Comparison for MU-MIMO• Fast Feedback channels
– Construction: Three 6x6 tiles make up a feedback LRU• Three 2x6 FMTs make up a feedback channel -- Three feedback channels in feedback LRU
– Base 16m codebook index is 6 bits– Each Primary Fast Feedback channel carries 6 bits
• A PFBCH LRU can carry 6-bit PMI for 3 users– Each Secondary Fast Feedback channel carries up to 12 bits
• A SFBCH LRU can theoretically carry 6-bit PMI for 6 users – Actually 3 users in practice since one user is allocated per SFBCH
• Analog feedback (AFB) channels– Construction: Three 6x6 tiles make up an AFB LRU
• For 4TX at BS: 6 AFB channels in an AFB LRU– An AFB LRU can carry eigenvectors for 6 users
• Conclusion for PMI Feedback vs AFB (4TX):– The same T-F “real estate” supports:
• MU-MIMO eigenvector feedback for 6 users• MU-MIMO PMI feedback for 3 users
– AFB has less overhead than the base codebook– However: Adaptive codebooks require the same PMI feedback as base codebook plus an
additional periodic long-term covariance feedback
• AFB requires half the overhead required by PMI feedback– And this ignores the additional periodic long-term covariance feedback required by the
adaptive codebooks!
17
System Simulation Setup (1/3)• System Layout:
– 19 cell (57 sector) classical layout; statistics in the center cell; – Cell radius = 866 m– 10 MS per sector; final statistics on about ~600 MS in the center cell– 10 MHz, full frequency reuse across cells/sectors
• Channel: – Pathloss model: 130.19 + 37.6*log10(R), where R in km + 10dB penetration loss
+ 2 dB cable loss– SCM channel details: PedB multipath profile, AS = 15 degrees, v = 3km/h, f =
2.5GHz; – Dominant interferers (path loss within 20dB of desired) are modeled as frequency
and spatially selective • BS:
– 4 Transmit antennas, 4 Receive antennas (power fair TX power allocation per antenna)
– TX power = 46 dBm; – ULA with half wave spacing; Array is calibrated– Sector antenna: parabolic (70 degrees 3 dB beamwidth), 17 dBi gain, 20 dB
front-to-back ratio• MS:
– 1 Transmit antenna, 2 receive antenna– MMSE receiver at the MS– UL CCH MS TX power = 23 dBm
• DL TX methods– SU-MIMO (realistic rank adaptation and PMI selection based on wideband
broadcast interference estimate at the MS on midamble)– MU-MIMO (realistic user grouping at the BS based on wideband PMI & CQI
feedback or eigenvector & CQI feedback, both based on midamble)• Wideband feedback
– Rank, CQI, PMI, and eigenvector calculation based on wideband midamble– Ideal DL channel estimation for Rank adaptation, CQI, PMI selection, and
eigenvector calculation– Rank-one feedback for MU-MIMO (for both PMI and eigenvector)
19
System Simulation Setup (3/3)• UL models for PMI & AFB:
– Ideal (PMI and AFB known at BS)– Realistic
• Includes UL CCH power control, realistic UL interference environment, realistic channel estimation and UL RX processing
• Codebook: MOT FFB proposal (PMI error rate between 0.5% and 1.5%)• AFB: Proposal in this set of slides is simulated (MUB matrix A for D=4 used by all MSs)
• UL Models for Rank and CQI feedback– Rank feedback for SU-MIMO is Ideal– CQI feedback is ideal
• Link adaptation (identical for PMI & AFB)– Realistic: based on wideband CQI feedback from the MS; 1 frame delay– Chase HARQ is modeled (max 3 attempts)
• Feedback delays: – 5 ms (1 frame) delay for both PMI and analog feedback– 50 ms (10 frames) delay for covariance matrix feedback to enable A-CBOOK
• A-CBOOK: Covariance matrix estimation and feedback is ideal, done at the midamble, interference-free, and quantized according to AWD
• UL power control SINR target: pre-CDM/MRC combining, per tone, per antenna– Varied for AFB: 0, 2, 4, 6 dB
• Link-to-system mapping: EESM• The effect of precoded interference seen on the DL data versus the midamble-
only interference seen during CQI and PMI/AFB calculations is captured
202020
System Level Results (16m EMD)
UL Model Ideal UL Complete UL
Feedback method
EigenvectorLTE
CBOOKLTE
A-CBOOKEigenvector
LTE CBOOK (~1%)
LTE A-CBOOK
(~1%)
SU-MIMO
105.9 100.0 106.8 98.8 (0 dB PC target)
100.2 (2 dB PC target)
101.9 (4 dB PC target)
102.0 (6 dB PC target)
99.7 106.4
MU-MIMO
166.7 136.2 146.9 136.2 (0 dB PC target)
139.8 (2 dB PC target)
149.7 (4 dB PC target)
152.1 (6 dB PC target)
135.9 146.5
Sector throughput (Mbps) – 16m EMD
12%
4%
Note: throughput results are normalized: SU-MIMO w/ ideal UL CBOOK = 100
4 Transmit Antennas
21
Summary• For SU-MIMO, analog feedback shows comparable performance to
base codebook and adaptive codebook methods• For MU-MIMO, analog feedback shows large gains over base
codebook– Gain over codebooks even with calibrated transmit chains– Much larger gains over codebooks with uncalibrated transmit chains– Analog feedback-based methods are not affected by uncalibrated
transmit chains• Analog feedback shows a gain over adaptive codebooks for MU-
MIMO– Adaptive codebooks are very promising and appear to provide gains
over standard-mode codebooks– Analog feedback still better than the adaptive codebooks
• Proposal for Analog feedback has half the overhead of PMI feedback– And doesn’t require the long-term covariance matrix feedback that is
required by the adaptive codebooks• Analog feedback can easily be included as an additional feedback
