3GPP TSG RAN WG1 Meeting #50bis
`Shanghai, China, October 08 – 12, 2007
`
`R1-074426
`
`Source:
`Title:
`Agenda Item:
`Document for:
`
`Panasonic
`Rank feedback in downlink MIMO
`6.4.5
`Discussion and Decision
`
`Introduction
`1.
`The general structure of precoding related feedback signaling was agreed at the St. Louis meeting [1]. We showed
`benefit of longer reporting interval (e.g. 20ms) for MIMO rank reporting with reliable transmission scheme [2]-[4].
`In this contribution we address considerations on rank feedback in downlink MIMO, which indicates the necessity
`for reliable rank suggestion with longer interval. In order to show the feasibility of proposed scheme, throughput
`performance results for different rank feedback intervals are presented, considering agreed codebook for 4tx at the
`Orlando meeting [5].
`
`2. UE feedbacks for SU-MIMO
`In order to support SU-MIMO, some additional feedbacks are necessary to compare with single antenna
`transmission or open-loop transmit diversity case, namely
`a) rank suggestion,
`b) precoding matrix indicator (PMI),
`c) CQI to support up to 2 codewords and
`d) Ack/Nack to support up to 2 codewords [6].
`Among them required bits for b) and c) depend on the corresponding contents of a) rank suggestion. For example,
`considering 2×2 (4×2) configuration there are two cases for rank selection.
` Rank 1
` CQI: same amount of bits as single stream transmission
` PMI: 3 (4) bits per
` order of 5 adjacent RBs or
` whole or subset of RBs
` Rank 2
` CQI: increased by 60-100% compared to single stream transmission [9][10]
` PMI: 2 (4) bits per
` order of 5 adjacent RBs or
` whole or subset of RBs
`Accordingly, total amount of feedback bits for each case is quite different. In order to minimize impact of this
`difference to the complexity of eNodeB reception, it is agreed that blind detection of CQI format should be avoided
`[7]. Then UE is necessary to be informed CQI format corresponding to rank. Considering this aspects, rank
`suggestion should be separated from PMI and CQI while latter two can be sent simultaneously as proposed in [6],
`and rank suggestion by UE is needed in advance to resource allocation by eNodeB for CQI/PMI reporting. Figure 1
`shows an example behavior.
`0) Uplink resource to suggest rank is allocated
`1) UE suggests rank for following reporting.
`2) eNodeB allocates uplink resources to the UE corresponding to the suggested rank.
`3) UE reports CQI/PMI using the allocated resources
`4) eNodeB transmits DL-SCH according to reported CQI/PMI.
`
`1
`
`APPLE 1006
`
`

`

`UE
`
`eNodeB
`
`Allocate uplink resources
`to suggest rank
`
`Suggest Rank for
`following reporting(s)
`
`Allocate uplink resources
`to corresponding feedback
`
`Transmit CQI/PMI using
`allocated resources
`
`Transmit DL-SCH
`according to reported
`CQI/PMI
`
`
`
`Figure 1 Example for reporting procedure
`
`
`Considering above kind of procedure, reliable transmission scheme for the rank suggestion is necessary in order for
`efficient resource allocation for uplink feedback and following downlink DL-SCH transmission. For the reliable
`transmission, that is possible for example, 1) transmission over PUSCH with CRC protection, or 2) transmission
`over PUCCH with similar protection level as Ack/Nack feedback. With separate coding between rank and PMI/CQI
`we can also use different reporting interval for those two, considering that longer reporting interval for rank than
`PMI/CQI brings better performance.
`
`As a conclusion of this section, following points should be noted here:
` Rank suggestion by UE is needed in advance to resource allocation by eNodeB for CQI/PMI reporting in
`order to avoid blind detection of CQI format, and
` Reliable transmission scheme for rank suggestion is necessary e.g. transmission over PUSCH with CRC
`protection or transmission over PUCCH with repetition among several sub-frames.
`
`
`3. Numerical analysis
`3.1. System simulation results
`In this section we demonstrate system simulation results in order to compare different rank reporting intervals
`among 1) within one radio-frame [5ms], 2) several radio-frames [20ms] and 3) ten(s) radio-frames [100ms], while
`fixed precoding reporting interval of 5ms. We assume either 1% or 10% of feedback error in addition to error free
`feedback on rank reporting, while either 1% or 10% of feedback error for precoding reporting (PMI error). These
`values are aligned with CQI block error rate target in [8]. Using system simulator we evaluated 4×2 antenna
`configuration with agreed codebook [5], assuming MMSE receiver. The other simulation parameters used to obtain
`the results are listed in appendix part.
`Figure 2 shows the results of different rank reporting intervals with 4×2 antenna configuration for uncorrelated
`typical urban (TU) channel, assuming the numbers of UEs in the sector as 5, 10 and 20, respectively. Figure 2
`indicates comparison of the user throughput of CDF 5% and sector throughput assuming 1% PMI error with 3km/h
`velocity.
`According to Figure 2, we can see the results with reporting interval within 20 ms have no significant performance
`loss compared to that of 5ms. This indicates 1% of rank reporting error has much impact for sector throughput than
`rank reporting delay if the delay is small, e.g. within 20ms.
`Table 1 summarizes results with 10UEs per sector for uncorrelated fading environments. The other results for
`correlated fading environments are described in appendix part.
`
`2
`
`

`

`In these cases, we also observe the results with reporting interval within 20ms have no significant performance loss
`compared to that of 5ms. This is aligned to that of 2×2 antenna configuration shown in [3] and appendix part.
`
`
`Rank reporting interval
`5ms 20ms
`100ms
`
`
`Rank reporting error
`without error
`with 1% error
`
`Figure 2 User throughput vs Sector throughput of 4×2 antenna configuration
`(PIM reporting interval = 5ms, PMI error = 1%, TU, i.i.d., v=3km/h)
`
`Table 1 Results for 4×2 TU i.i.d. (v=3, 15km/h)
`10%
`
`
`
`1%
`
`10%
`
`1%
`
`error-free
`
`1%
`
`error-free
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`2.153 2.158 2.132 2.250 2.246 2.228 2.261 2.256 2.239 2.430 2.427 2.407 2.444 2.438 2.420
`
`0.070 0.071 0.070 0.073 0.074 0.073 0.074 0.074 0.074 0.079 0.079 0.079 0.080 0.080 0.080
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.546 1.547 1.557 1.585 1.583 1.594 1.589 1.588 1.598 1.651 1.650 1.662 1.656 1.655 1.667
`
`0.051 0.051 0.052 0.053 0.053 0.053 0.053 0.053 0.053 0.055 0.055 0.055 0.055 0.055 0.056
`
`
`
`4x2 TU i.i.d.
`
`PMI
`error
`Rank
`error
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`
`3
`km/h
`
`15
`km/h
`
`3
`
`

`

`
`
`3.2. Link simulation results
`We also evaluated 4×2 antenna configuration using link simulator assuming almost similar parameters to system
`evaluation listed in appendix part.
`According to Figure 3 (a) and (b) we can see the results with reporting interval within 20ms bring no significant
`performance loss compared to that of 5ms for both 3 and 15 km/h velocities as well as system evaluation results.
`This is aligned to that of 2×2 antenna configuration [3] as well.
`
`
`4x2, TU 15km/h (tx 0.5, rx 0.0)
`
`5ms
`10ms
`20ms
`50ms
`100ms
`
`8.0
`
`7.0
`
`6.0
`
`5.0
`
`4.0
`
`3.0
`
`2.0
`
`1.0
`
`0.0
`
`Spectral efficiency [bps/Hz] a
`
`4x2, TU 3km/h (tx 0.5, rx 0.0)
`
`5ms
`10ms
`20ms
`50ms
`100ms
`
`8.0
`
`7.0
`
`6.0
`
`5.0
`
`4.0
`
`3.0
`
`2.0
`
`1.0
`
`0.0
`
`Spectral efficiency [bps/Hz] a
`
`25
`
`0
`
`5
`
`20
`
`25
`
`
`
`0
`
`5
`
`20
`
`15
`10
`15
`10
`Geometry [dB]
`Geometry [dB]
`
`(a) 3km/h (b) 15km/h
`Figure 3 Throughput of 4×2 configuration
`
`
`4. Conclusion
`In this document, we addressed considerations on rank feedback in downlink MIMO and showed the throughput
`performance among different rank reporting intervals for SU-MIMO. For rank feedback in downlink MIMO, we
`propose longer reporting interval (e.g. 20ms) for MIMO rank reporting with reliable transmission scheme. The
`reliable transmission is possible for example, 1) transmission over PUSCH with CRC protection, or 2) transmission
`over PUCCH with similar protection level as Ack/Nack feedback. Besides, according to the system simulation
`results 20ms rank reporting interval have negligible performance loss compared to 5ms interval case. On the other
`hand 10% of feedback error on rank reporting causes a certain performance loss. We can see similar tendency of
`rank reporting interval from link level results as well.
`
`References
`[1]
`R1-071228, LTE MIMO session chairman, “LTE MIMO AH Summary”
`[2]
`R1-072085, Panasonic, “Evaluation of slow rank adaptation”
`[3]
`R1-072806, Panasonic, “Rank feedback consideration”
`[4]
`R1-073633, Panasonic, “Rank feedback in downlink MIMO”
`[5]
`R1-073206, MIMO Ad-hoc session, “Text Proposal for TS36.211 for 4-Tx Antenna SU-MIMO Codebook”
`[6]
`R1-073423, Texas Instruments, “Design Aspects of UE Feedback”
`[7]
`R1-073858, Ericsson, Alcatel-Lucent, CATT, Freescale Semiconductor, Huawei, Icera Semiconductor,
`Interdigital, IPWireless, LGE, Mitsubishi, Motorola, Nextwave, Nokia, Nokia Siemens Networks, Nortel,
`NTT DoCoMo, Panasonic, Philips, Qualcomm Europe, Samsung, Sharp, Texas Instruments, ZTE , “Way
`forward for CQI reporting”
`R1-071839, RAN1, “LS on target quality on L1/L2 control channel”
`[8]
`R1-073489, LG Electronics, “Reduction of MIMO CQI reporting in spatial domain”
`[9]
`[10] R1-073632, Panasonic, “CQI quantization for SU-MIMO”
`
`
`4
`
`

`

`Appendix
`<Simulation parameters for system level evaluation>
`Detailed simulation parameters for system level evaluation are listed here.
`Table A-1 Macro-cell system simulation parameters
`Parameter
`Assumption
`Cellular Layout
`Hexagonal grid, 19 cell sites, 3 sectors per site
`Inter-site distance
`500m
`Frequency Reuse
`1
`Carrier Frequency / Bandwidth
`2 GHz
`Antenna configuration
`4x2, 2x2
`Channel model
`Uncorrelated channel:
`Typical Urban
`Correlated channel:
`SCM-C
`3, 15 km/h
`46dBm (1Antenna) – 10MHz carrier
`Users dropped uniformly in a cell of 3R radius
`Chase combining, Non-adaptive, Asynchronous
`6 TTI (6ms)
`5
`10%
`
`UE speed
`Total BS TX power (Ptotal)
`Macro-diversity
`HARQ
`Delay between retransmissions
`Maximum retransmissions
`Target PER
`
`
`
`Table A-2 OFDMA system simulation parameters
`Parameter
`Assumption
`
`TTI duration
`Transmission BW
`Usable sub-carriers
`CP Length
`Number of OFDM symbols per sub-frame
`
`1.0ms
`10MHz
`600
`Short
`10 (data) + 4 (control+pilot)
`
`Table A-3 Scheduling parameters
`Parameter
`Assumption
`Scheduling granularity
`900 kHz bandwidth (5RBs)
`60 sub-carriers x 14 symbols
`71.5%
`5 TTI (5ms)
`Proportional Fair
`
`Useful symbol rate
`CQI feedback delay
`Scheduler
`
`Qpsk(R = 1/ 8)
`Qpsk(R = 2/ 5)
`16Qam(R = 2/ 5)
`64Qam(R = 1/ 2)
`64Qam(R = 7/10)
`
`Qpsk(R = 1/ 5)
`Qpsk(R = 1/ 2)
`16Qam(R = 1/ 2)
`64Qam(R = 11/ 20)
`64Qam(R = 3/ 4)
`
`Qpsk(R = 1/ 4)
`Qpsk(R = 3/ 5)
`16Qam(R = 3/ 5)
`64Qam(R = 3/ 5)
`64Qam(R = 4/ 5)
`
`Qpsk(R = 1/ 3)
`16Qam(R = 7/ 20)
`16Qam(R = 2/ 3)
`64Qam(R = 2/ 3)
`64Qam(R = 5/ 6)
`
`1.00E+00
`
`1.00E-01
`
`1.00E-02
`
`PER
`
`1.00E-03
`
`-6
`
`-4
`
`-2
`
`0
`
`2
`
`4
`
`8
`6
`SNR [dB]
`
`10
`
`12
`
`14
`
`16
`
`18
`
`20
`
`
`
`Figure A-1 PER curves used for MCS selection and throughput calculation
`
`
`
`
`
`
`
`5
`
`

`

`Figure A-2 CDF of Geometry
`
`
`<System level results for correlated fading environments>
`Table A-4 Results for 4×2 SCM-C (v=3, 15km/h)
`10%
`
`
`
`1%
`
`10%
`
`1%
`
`error-free
`
`1%
`
`error-free
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`2.110 2.119 2.102 2.211 2.211 2.199 2.222 2.221 2.210 2.394 2.395 2.381 2.408 2.407 2.395
`
`0.068 0.068 0.068 0.071 0.071 0.071 0.071 0.071 0.071 0.077 0.077 0.077 0.077 0.077 0.077
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.653 1.655 1.655 1.703 1.701 1.703 1.709 1.706 1.708 1.787 1.785 1.788 1.794 1.791 1.793
`
`0.054 0.054 0.054 0.055 0.056 0.056 0.056 0.056 0.056 0.058 0.058 0.059 0.059 0.059 0.059
`
`4x2 SCM-C
`
`3
`km/h
`
`15
`km/h
`
`PMI
`error
`Rank
`error
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`
`
`6
`
`

`

`<System level results for 2×2>
`
`Table A-5 Results for 2×2 TU i.i.d. (v=3, 15km/h)
`10%
`
`1%
`
`10%
`
`1%
`
`error-free
`
`1%
`
`error-free
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.929 1.941 1.921 1.995 1.989 1.961 2.002 1.995 1.965 2.096 2.091 2.064 2.103 2.095 2.067
`
`0.066 0.067 0.067 0.068 0.069 0.068 0.069 0.069 0.069 0.072 0.072 0.072 0.072 0.072 0.072
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.390 1.400 1.399 1.427 1.423 1.419 1.430 1.425 1.421 1.476 1.474 1.471 1.480 1.476 1.472
`
`0.043 0.044 0.044 0.044 0.044 0.044 0.044 0.044 0.044 0.046 0.046 0.046 0.046 0.046 0.046
`
`
`Table A-6 Results for 2×2 SCM-C (v=3, 15km/h)
`10%
`
`1%
`
`10%
`
`1%
`
`error-free
`
`1%
`
`error-free
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.896 1.914 1.902 1.961 1.960 1.941 1.967 1.965 1.946 2.067 2.068 2.050 2.074 2.073 2.054
`
`0.064 0.064 0.065 0.066 0.066 0.066 0.066 0.066 0.066 0.070 0.070 0.069 0.070 0.070 0.070
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`5
`
`20
`
`100
`
`1.498 1.504 1.504 1.533 1.526 1.523 1.536 1.529 1.525 1.573 1.568 1.565 1.576 1.569 1.566
`
`0.051 0.051 0.052 0.052 0.052 0.052 0.052 0.052 0.052 0.053 0.054 0.054 0.054 0.054 0.054
`
`2x2 TU i.i.d.
`
`PMI
`error
`Rank
`error
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`
`3
`km/h
`
`15
`km/h
`
`3
`km/h
`
`15
`km/h
`
`2x2 SCM-C
`
`PMI
`error
`Rank
`error
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`Rank
`reporting
`interval [ms]
`Sector
`throughput
`[bps/Hz]
`5% user
`throughput
`[bps/Hz]
`
`
`7
`
`

`

`<Simulation parameters for link level evaluation>
`Detailed simulation parameters for link level evaluation are listed here.
`Table A-7 Link simulation assumptions
`10MHz
`0.5 ms
`1.0 ms
`15kHz
`15.36 MHz
`12 sub-carriers
`5 (60 sub-carriers)
`1024
`601 (DC sub-carrier is null)
`7
`Turbo code, R=1/3
`20 levels
`[QPSK, R=1/8] [QPSK, R=1/5] [QPSK, R=1/4] [QPSK, R=1/3]
`[QPSK, R=2/5] [QPSK, R=1/2] [QPSK, R=3/5]
`[16QAM, R=7/20] [16QAM, R=2/5] [16QAM, R=1/2]
`[16QAM, R=3/5] [16QAM, R=2/3]
`[64QAM, R=1/2] [64QAM, R=11/20] [64QAM, R=3/5]
`[64QAM, R=2/3] [64QAM, R=7/10] [64QAM, R=3/4]
`[64QAM, R=4/5] [64QAM, R=5/6]
`4x2
`TU with Kronecker extension
`MMSE channel estimation
`28.5% (14.3% for pilot and 14.2% for signaling)
`Max-Log-MAP with 8 iterations
`5 ms delay without feedback error
`5 / 10 / 20 / 50 / 100 ms
`Sub-band (continuous 5RBs) bases RR
`Non-blanking based IR with maximum 4 transmission
`(non-adaptive, synchronous in time and freq. domain: use same
`RBs with a period of 6ms)
`
`
`<Link level results for 2×2 on [3]>
`
`2x2, TU 15km/h (tx 0.5, rx 0.0)
`
`5ms
`10ms
`20ms
`50ms
`100ms
`
`6.0
`
`5.0
`
`4.0
`
`3.0
`
`2.0
`
`1.0
`
`0.0
`
`Spectral efficiency [bps/Hz] a
`
`2x2, TU 3km/h (tx 0.5, rx 0.0)
`
`5ms
`10ms
`20ms
`50ms
`100ms
`
`6.0
`
`5.0
`
`4.0
`
`3.0
`
`2.0
`
`1.0
`
`0.0
`
`Spectral efficiency [bps/Hz] a
`
`0
`
`5
`
`
`
`20
`
`25
`
`0
`
`15
`10
`15
`10
`Geometry [dB]
`Geometry [dB]
`
`(a) 3km/h (b) 15km/h
`Figure A-3 Throughput of 2×2 configuration
`
`5
`
`20
`
`25
`
`
`
`Transmission BW
`Slot duration
`Sub-frame duration
`Sub-carrier spacing
`Sampling frequency
`RB size
`Number of RBs used
`FFT size
`Number of occupied sub-carriers
`Number of OFDM symbols per slot
`Channel coding
`
`Modulation and coding rate
`
`Number of antennas
`Channel environments
`Channel estimation
`Pilot and signaling overhead
`FEC Decoder algorithm
`CQI reporting delay
`Rank reporting delay
`Frequency scheduling
`
`HARQ
`
`8
`
`