Performance Evaluation of Codebooks Proposed for IEEE m Amendment IEEE Presentation Submission Template (Rev. 9) Document Number: IEEE C80216m-09_0344 Date Submitted: Source: David Mazzarese, Bruno Clerckx, Kwanhee Roh, Wang Zhen, Heewon Keun Chul Hwang, Sungwoo Park, Soon-Young Yoon, Hokyu Choi, Jerry Pi Kaushik Josiam, Sudhir Ramakrishna, Farooq Khan Samsung Electronics Venue: IEEE m Session#59, San Diego, US IEEE m-08/053r1, “Call for Contributions for P802.16m Amendment Text Proposals”. Topic: “DL MIMO and UL MIMO”. Base Contribution: IEEE C80216m-09_0344 Purpose: Discussion and approval Notice: This document does not represent the agreed views of the IEEE 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 Patent Policy: The contributor is familiar with the IEEE-SA Patent Policy and Procedures: and. Further information is located at and.

Introduction Several codebooks have been proposed in the base contribution C80216m-09_0279. This contribution provides supporting simulations and analysis of base codebooks.

Codebook-based feedback Codebooks are used in uplink feedback to the BS for supporting downlink precoding 3 codebook-based feedback modes in SDD –Standard: base codebook –Adaptive: correlation matrix transformation –Differential: differential PMI feedback

Base Codebook Analysis

Base Codebook Candidates ReferenceAuthorLabel used in figures e802.16e16e_3bit e802.16e16e_6bit C80216m-09_0279David Mazzarese09/0279_4bit C80216m-09_0279David Mazzarese09/0279_6bit C80216m-09_0056Guangjie Li09/0056_4bit C80216m-09_0056Guangjie Li09/0056_6bit C80216m-09_106r1Yang Tang09/0106_6bit C802.16m-08_983 C80216m-MIMO-08/69 Jaewan Kim08/983_4bit C80216m-08_1101 C80216m-MIMO-08/74 Bishwarup Mondal08/1101_4bit C80216m-MIMO-08_067Shaohua li08/067_6bit C80216m-08/916 C80216m-08/1264r1 Ron Porat08/916_4bit

Measures of Codebook Goodness MeasureDescription ThroughputFirst and foremost measure Feedback overheadCodebook size (number of bits) DFT structureBest for calibrated correlated linear arrays Block-diagonal matricesAdapted for dual polarized arrays at BS Full nested propertyRanks 2, 3 and 4 matrices are composed of rank 1 precoders: CQI computation complexity reduction Constant modulus matrix elements Good for power amplifier transmit power balance, good for PAPR in precoded systems QPSK alphabetCQI computation complexity reduction by avoiding numerous complex multiplications Avoidance of rank deficiency weakness Any combination of 2 to Nt rank 1 precoders produces an invertible matrix: desirable for simple implementation complexity of ZFBF

Comparison of 4Tx Codebooks 09/ e09/ /010608/98308/1101 Feedback overhead 4 bits6 bits3/6 bits4 bits6 bits 4 bits Performance (SLS) C: good U: good C: good U: good C: bad U: good C: good U: good C: good U: good C: good U: good C: good U: bad C: bad U: bad DFT structureYes NoYes PureYes Block-diagonal matrices No YesNo Full nested property Full No Yes (rank 1 and 2) YesNo Constant modulus YesNo Yes QPSK alphabetNo Yes Avoidance of rank deficiency Yes No YesNo C: correlated channels U: uncorrelated channels

Performance Evaluation

Simulation Environments SU MIMO SLS in DL 4x2 –ULA: uncorrelated, correlated channels MU MIMO (ZFBF) SLS in DL 4x2 –ULA: uncorrelated, correlated channels

SU MIMO SLS in DL 4x2 Uncorrelated Channel (4 lambdas)

SU MIMO SLS in DL 4x2 Correlated Channel (1/2 lambda)

SU MIMO SLS in DL 4x2 16e codebooks lose 6% and 11% throughput in correlated channels in reference to the best 6-bit codebook DFT-based codebooks are robust in all scenarios 6-bit DFT-based codebooks are as good or better than 16e in uncorrelated channels All 6-bit DFT-based codebooks are within 0.5% of each other All 4-bit DFT-based codebooks are within 1% of each other The 4-bit DFT-based codebooks lose only 1% and 3% to the 6-bit DFT- based codebooks in correlated and uncorrelated channels, respectively The transformation brings the performance of small codebooks to the same level as the 6-bit DFT-based codebooks in correlated channels The transformation is ineffective in uncorrelated channels

SU MIMO SLS in DL 4x2  4-bit codebook seems like a reasonable choice, since it requires lower computational complexity and feedback overhead than a 6-bit codebook

MU MIMO SLS in DL 4x2 Uncorrelated Channel (10 lambdas)

MU MIMO SLS in DL 4x2 Correlated Channel (1/2 lambda)

MU MIMO SLS in DL 4x2 16e codebooks lose 1% and 22% throughput in correlated channels in reference to the best 6-bit codebook DFT-based codebooks are robust in all scenarios 6-bit DFT-based codebooks are as good or better than 16e in uncorrelated channels All 6-bit DFT-based codebooks are within 1% of each other in uncorrelated channels All 6-bit DFT-based codebooks are within 1% of each other in correlated channels, except 09/0056_6bit which loses 4.5% to 09/0279_6bit All 4-bit DFT-based codebooks are within 1% of each other The 4-bit DFT-based codebooks lose only 7% and 8% to the 6-bit DFT-based codebooks in correlated and uncorrelated channels, respectively The transformation allows the performance of 4-bit codebooks to exceed the 6- bit DFT-based codebooks by 3% in correlated channels The transformation is ineffective in uncorrelated channels

MU MIMO SLS in DL 4x2  The average sector throughput with MU MIMO in correlated channels is about 1.5 times higher than in uncorrelated channels. This fact stresses the importance of 1.Optimizing the codebook in the correlated channel 2.Calibrating the antenna array at the BS to avoid random phase effects to benefit from the DFT-based structure  Thus a 4-bit codebook seems like a reasonable choice, since it requires lower computational complexity and feedback overhead than a 6-bit codebook

Feedback Overhead Penalty of 6-bit codebook vs. 4-bit codebook –It depends on feedback channel design CQICH or feedback header used for PMI feedback? How many PMIs carried in one CQICH? Subband info carried in same CQICH as PMI? How many best subbands? –It also depends on the number of users that feedback PMI More knowledge of feedback channel design and analysis of feedback procedure is necessary before finally choosing between a 4-bit and a 6-bit base codebook

Appendix Simulation Assumptions

Number of Antennas 2 transmitter, 2 receiver [2Tx, 2Rx] 4 transmitter, 2 receiver [4Tx, 2Rx] 4 transmitter, 4 receiver [4Tx, 4Rx] Antenna configuration ULA: 0.5 lambda; 4 lambda, 10 lambda Split Linear Array, Dual Polarized Array MIMO Scheme 1.Closed-loop single user with dynamic rank adaptation 2.Zero-forcing multiple user MIMO Schedule from 1 to 2 users dynamically based on the same rank-1 PMI feedback. No SU/MU mode adaptation. Channel ModelModified Ped-B 3km/h Channel correlation Scenario 1. Uncorrelated Channel : 4 lambda antenna spacing, angular spread of 15 degrees 2. High correlated channel: 0.5 lambda antenna spacing, angular spread of 3 degree PAPR1. No constraint on per-antenna power imbalance 2. Limitation of per-antenna power imbalance by scaling in every subframe Antenna Calibration1.Ideal antenna calibration (mandatory) 2.Uncalibrated antennas (optional)  Random phase on each transmit antenna + Random delay between each pair of adjacent transmit antennas (uniformly distributed between 0 and N samples)  Fixed for one drop

OFDM parameters10 MHz (1024 subcarriers) OFDM symbols per subframe6 PermutationLocalized Number of total RU in one subframe48 Scheduling Unit Whole band (48 PRUs) 12 subbands 1 subband = 4 consecutive PRUs 1 PMI and 1 CQI feedback per subband Number of RU for PMI and CQI calculation 4 which is same as in IEEE e CQI, PMI feedback periodEvery 1 frame (5ms) Feedback delay1 frame (5ms) Link Adaptation (PHY abstraction) QPSK 1/2 with repetition 1/2/4/6, QPSK 3/4, 16QAM 1/2, 16QAM 3/4, 64QAM 1/2, 64QAM 2/3, 64QAM 3/4, 64QAM 5/6

HARQ Chase combining, non-adaptive, asynchronous. HARQ with maximum 4 retransmissions, 4 subframes ACK/NACK delay, no error on ACK/NACK. HARQ retransmission occurs no earlier than the eighth subframe after the previous transmission. SchedulingNo control overhead, 12 subbands of 4 PRUs each, latency timescale 1.5s MIMO receiverLinear Minimum Mean Squared Error (LMMSE) Data Channel EstimationPerfect data channel estimation Feedback Channel MeasurementPerfect feedback channel measurement Cellular Layout Hexagonal grid, 19 cell sites, wrap-around, 3 sectors per site Distance-dependent path lossL= log 10 (.R), R in kilometers Inter site distance1.5km Shadowing standard deviation8 dB Antenna pattern (horizontal) (For 3-sector cell sites with fixed antenna patterns) = 70 degrees, A m = 20 dB Users per sector10 (EMD) Scheduling CriterionProportional Fair (PF for all the scheduled users) Feedback channel error rateNo error

23/#NN Power fluctuation among antennas Constant modulus property –Definition: Every elements of codebook vector has same magnitude –Good for per-antenna peak power limit –DFT-based codebooks have a constant modulus property, while 16e-based do not Total power limitation Per-ant. Peak power limitation Sum power is limited to 20W Per-ant power limited to 5W Power adjustment subframe by subframe