New Radio Access Technology for 5G

Transcription

1 New Radio Access Technology for 5G Jen Ming Wu ( 吳仁銘 ) Inst. of Communications Engineering Dept. of Electrical Engineering National Tsing Hua University Hsinchu, Taiwan 1 September, 2016

2 Outline Introduction 5G New Radio 5G Objectives 4G LTE based on OFDMA Basics of OFDM New Waveforms for 5G New Radio as a 3GPP Study Item Candidate Waveforms Non Orthogonal Multiple Access (NOMA) Power Domain Code Domain Summary 2

3 Timeline of New Radio in 3GPP The new RAT shall be inherently forward compatible. The normative specification would occur in two phases: Phase I: address a more urgent subset of the commercial needs Phase II: address all identified use cases and requirements for IMT 2020 submission Three study items approved in RAN plenary: Channel model above 6GHz (TR V0.4.0, 3GPP RAN1 Meeting #85) Usage Scenario and Requirements (TR V0.3.0, 3GPP RAN1 Meeting #85) New Radio Access Technology :Physical Layer Aspects (TR V0.0.2, R ) 3

4 5G Objectives Defined in 3GPP 3GPP TR : Study on Scenarios and Requirements for Next Generation Access Technologies. (Dec. 2015) Considerations Performance Implementation complexity Latency Flexibility embb (Enhanced Mobile Broadband): Low latency Higher spectral efficiency/throughput mmtc(massive machine type communications): Improved link budget Low device complexity Long device battery life High density device deployment URLLC(Ultra reliable low latency communications): High reliability (low packet error rate) Low latency 4

5 Numerologies Agreements in RAN1#84bis <R > Largest component carrier bandwidth not smaller than 80 MHz for at least one numerology is supported Waveform is based on OFDM Agreements in RAN1#85 <R and R > Multiplexing different numerologies within a same NR carrier bandwidth (from the network perspective) is supported FDM and/or TDM multiplexing can be considered RAN1 concludes on alternative 1 (15 khz) as the baseline design assumption for the NR numerology RAN1 concludes on scale factors N =2 n for subcarrier spacing as the baseline design assumption for the NR numerology 5 Source: R /QualComm

6 Agreement in RAN1#84bis <R > Frame Structure Study flexible/dynamic TDD, including both downlink and uplink transmissions in the same subframe interval Agreement in RAN1#85 <R > A time interval X which can contain one or more of the following DL transmission part Guard UL transmission part NR design should strive at least to enable the possibility for Corresponding ack reporting shortly (in the order of X µs) after the end of the DL data transmission Corresponding uplink data transmission shortly (in the order of Y µs) after reception of UL assignment Note: may depend on e.g. UE capability/category, payload size, etc FFS: X and Y in the order of a few tens of or hundreds of micro sec is feasible 6

7 Orthogonal Frequency Division Multiple Access (OFDMA) 4G LTE based on OFDMA Motivation for a New Radio in 5G New application demands in 5G may drive the need to develop new radio access technologies that are not necessarily backward compatible to 4G. 7

8 OFDMA Advantages: Strict orthogonality High BW efficiency Low complexity in equalization. Effective ISI elimination with use of cyclic prefixes (CP). Disadvantages: Sensitivity to CFO (ICI). Stringent requirements for synchronization. BW efficiency loss due to guard band Motivation for a New Radio in 5G New application demands in 5G may drive the need to develop new radio access technologies that are not necessarily backward compatible to 4G. 8

9 A Unified Frame Structure for 5G A unified frame structure to incorporate users with diverse application requirements and trac characteristics [5GNOW IEEE ComMag Feb 14]. Type I: high rate data (video, etc.) Type II: high rate data with advanced receiver processing (e.g., CoMP). Type III: Sporadic MTC trac with relaxed synchronization requirements. Type IV: Sporadic asynchronous MTC trac; ultra low latency. 9

10 Orthogonal Frequency Division Multiplexing The available channel bandwidth is divided into N subchannels, each of bandwidth f, i.e, BT Nf Assign a subcarrier signal for each subchannel. Suppose each subcarrier is modulated with M ary QAM symbols. The signal on the kth subcarrier is B T s k 2 2 ( t) Aki cos(2f kt) Akq sin(2f kt), k 0,1,... N 1 T T 2 Re{ X T where X k A k ki e j 2f t k ja } kq, A ki A kq { 1, 3,... ( M 1)} 10

11 Orthogonal Frequency Division Multiplexing 11 T t e T t t f j k k 0, 2 ) ( 2 j k j k dt t t j T k 0, 1, ) ( ) ( * 0 X k s f T s s B T f / 1

12 Orthogonal Frequency Division Multiplexing 2 j k t = e t T T j2p fkt Time domain ( ), 0 ( ) Freq domain: fk( f ) = 2T sin cé2p f - fk Tù ë û 1 D f = NT s 12

13 Implementation of OFDM with IDFT/DFT Direct implementation requires N analog RF frontends! The cost is too expensive and prevents the OFDM realization for 20 years since the birth of concept of OFDM. It can be shown that the OFDM processing is mathematically equivalent to the IDFT/DFT. The IDFT/DFT processing can be realized in digital baseband with low cost. f Let k s f c k ( t) kf Re{ Re{ 2 j X ke T 2 X ke T 2f t k } e j 2kft j 2f t, k 0,..., N 1 c } 13

14 Implementation of OFDM with IDFT/DFT X k ( t) 2 X T k e j 2kft, k 0,..., N 1 j2 fct The passband signal is sk( t) Re{ Xk( t) e }, k 0,..., N 1 N-1 N-1 2 xt () = å X () t = å Xe k T k k= 0 k= 0 j 2pkDft N1 N1 () () Re () Re c st sk t Xk te xte () k0 k0 j2 ft j2 ft c 14

15 Implementation of OFDM with IDFT/DFT The discrete-time representation of x(t) at t=nt s is x( nt s ) x[ n] N 1 k 0 2 T X k e j 2kf nt s Recall: N 1 k j 2k ( NT X ke T s ) nt s 1 2 N - j2 pkn / N Xe k k= 0 = å T The baseband signal relation is equivalent to DFTpairs, DFT xn [ ] X k IDFT where k is index of subcarriers, and n is index of time. 15

16 X k Implementation of OFDM with IDFT/DFT x[ n] becomes IDFT / DFT pairs The OFDM Tx can be represented by IDFT The OFDM Rx can be represented by DFT X x[0] 0 x[ N 1],, x[0] X N 1 X 0 N -IDFT X x[ N 1] N 1 X x[0] 0 Add CP D/A e j 2 f c t X N 1 X 0 X N 1 Freq domain signals N -DFT x[ N 1] Remove CP Time domain signals A/D e j 2 c 16 f t

17 Freq Time Representation of OFDM Signal 17

18 OFDM Signaling over Multipath Channel Issues: OFDM requires accurate synchronization With a frequency offset between the transmitter and the receiver, the orthogonality of subcarriers would be destroyed, causing inter carrier interference (ICI). Strict (frequency) synchronization is needed by OFDMA for a satisfactory performance. 18

19 Outline Introduction 5G New Radio 5G Objectives 4G LTE based on OFDMA Basics of OFDM New Waveforms for 5G New Radio as a 3GPP Study Item Candidate Waveforms Non Orthogonal Multiple Access (NOMA) Power Domain Code Domain Summary 19

20 Fundamental physical layer signal structure for new RAT Waveform based on OFDM, with potential support of nonorthogonal waveform and multiple access FFS: other waveforms if they demonstrate justifiable gain Basic frame structure(s) Channel coding scheme(s) Radio interface protocol architecture and procedures Radio Access Network architecture, interface protocols and procedures, [DOCOMO, RP , March 2016] 20

21 New Waveforms for 5G FBMC (Filter bank based multi carrier) A filter is applied for each sub carrier No cyclic prefix, removal of ISI is usually complicated Extension to MIMO is usually non trivial. OOB leakage suppression is reduced with PA nonlinearity 0-10 FBMC:24 tones,60 symbols per run,1000 runs FBMC: clip at 8dB FBMC: no clipping WOLA: clip at 8dB WOLA: no clipping -20 db Normalized freq [1/T] M500/ICL

22 New Waveforms for 5G UFMC (Universal Filtered Multi Carrier) Basic idea: Applying a filter on a per RB basis, thereby greatly reducing the filter order (esp. compared to FBMC). Transmitter Al-Luc (Nokia) proposed UFMC in late 2013, using a filter length of only 1/3 or 1/4 OFDM symbol duration. Subject to ISI due to the lack of CP Receiver

23 Filtered OFDM (f OFDM) New Waveforms for 5G Spatial case for UFDM with longer filter length Still has CP thus simpler equalizer than UF OFDM Has ISI, but claims the impact is small Huawei proposed f OFDM in 2015, using a filter length of 0.5 OFDM symbol duration. CP-OFDM signal: CP Symbol N-1 CP Symbol N CP Symbol N+1 f-ofdm signal after filtering: CP Symbol N-1 CP Symbol N CP Symbol N+1 (c) f-ofdm filtering illustration M500/ICL

24 New Waveforms for 5G CP OFDM with WOLA (W OFDM) Weighted Overlap and Add (WOLA) Practical implementations using time domain windowing Better OOB suppression then CP OFDM Efficient CP length is reduced. subject to ISI under non flat channels WOLA: cp=0.1 txwola=0.078,12 tones,60 symbols per run,1000 runs CP-OFDM: no clipping WOLA: clip at 6dB WOLA: clip at 8dB WOLA: no clipping db Normalized freq [1/T] Tx-WOLA Rx-WOLA

25 DFT OFDM New Waveforms for 5G Precoder based waveform The fixed length CP is replaced by variable length zero tail, based on channel delay spread on a per user basis. Extra signaling overhead. The OOB leakage can be reduced due to the zero padding. A slight increase in PAPR due to the decreased average power Tx Rx [Qualcomm R ]

26 New Waveforms for 5G Generalized Frequency Division Multiplexing (GFDM) Replaces IFFT in OFDM with a precoding matrix Divide an OFDM symbol into subsymbols and subcarriers. Applies CP as a guard interval just like in CP OFDMA or SC CP. Low out of band emission (with a well chosen prototype filter). Originally proposed for cognitive radio [Fettweis, 2009]. G.Fettweis, M.Krondorf and S.Bittner, GFDM Generalized Frequency Division Multiplexing, IEEE VTC, Spring,

27 New Waveforms for 5G Generalized Frequency Division Multiplexing (GFDM) Disadvantages Amplifies noise at the receiver. Larger complexity than OFDM. Not orthogonal between subcarriers and subsymbols Not known how multiple access (MA) can be done (No such a thing as GFDMA). 27

28 Outline Introduction 5G New Radio 5G Objectives 4G LTE based on OFDMA Basics of OFDM New Waveforms for 5G New Radio as a 3GPP Study Item Candidate Waveforms Non Orthogonal Multiple Access (NOMA) Power Domain Code Domain Summary 28

29 Introduction of Multiple Access In general, different users (UE s) data can be multiplexed in time, frequency, code, spatial and power domains TDM, FDM, OFDMA, CDM and MU MIMO(perfect CSI for precoding) Freq Freq Freq B B T (a) TDMA Time T (b) FDMA Time Freq B Freq B Spatial Beam 5 Spatial Beam 4 Spatial Beam 3 Spatial Beam 1 Spatial Beam 2 Spatial Beam 1 (c) RSMA T Time (d) SDMA T Time Multiple access in New RAT are divided into two parts for discussion: OMA (waveform) OFDMA based waveform + SDMA may be the baseline OMA scheme in NR» Frequency, time and spatial domains are used Non OMA (Multiple Access) Most of Tdocs focus on power and code domain multiplexing 29

30 OMA and NOMA x h 2 BS h 1 y 1 User A y 2 User B Shared resource block for users Wide transmission bandwidth per user Narrow transmission bandwidth per user No inter user (intra cell) interference 30

31 Non orthogonal Multiple Access (NOMA) Power Domain Y. Saito el al, Non Orthogonal Multiple Access (NOMA) for Cellular Future Radio Access Docomo Spring VTC

32 Multiple Access in New RAT Three types of multiple access schemes are proposed: Power domain: NOMA power domain only (MUST)<R / NTT DoCoMo> Underlay common control <R /Idaho National Laboratory> Code domain: Multi user shard access(musa) <R /ZTE> Resource spread multiple access (RSMA)<R /Qualcomm> Non orthogonal coded multiple access (NCMA) <R /LG> Low code rate spreading /Frequency domain spreading <R /Intel> Low code rate and signature based shared access (LSSA)<R /ETRI> Non orthogonal coded access (NOCA)<R /Nokia> Hybrid power and code domain: Sparse code multiple access (SCMA)<R /Huawei> Pattern defined multiple access (PDMA)<R /CATT> Interleave Grid Multiple Access (IGMA)<R /Samsung> Multiple Access may be discussed until RAN1#86bis (2016/10) 32

33 Power Domain Power Domain Multiple Access Schedule the near UE1 and the far UE2 in the same time frequency resource Due to near far effect, the power of received signal from UE1 is always higher than the received power of UE2 At BS side, SIC should be applied Pros: No big change in the physical layer at the transmitter(ue) side Cons: Heavy dependency on the power imbalance between scheduled UEs Complex user grouping is required (Limit the scheduling flexibility) Can not provide enough protection especially for small packets 33

34 Power Domain Multiple Access Fettweis Y. Saito el al, Non Orthogonal Multiple Access (NOMA) for Cellular Future Radio Access Docomo Spring VTC

35 Capacity Achieving NOMA Multiuser Superposition Transmission (MUST) with successive interference cancellation (SIC) has been considered has been considered to enable NOMA in OFDMA Take the two user case for example Downlink transmission from BS to UE1 & UE2 QPSK modulation used by both UEs BS x 1 [n]+x 2 [n] UE2 UE1 High Received SNR Low x 1 [n] x 2 [n] x 1 [n]+x 2 [n] 35

36 Capacity Achieving NOMA Downlink Multiuser Superposition Transmission (MUST) with successive interference cancellation (SIC) Let the transmit power P P 1, P, P 1 2 Transmit x x1 1 x2, 0.5, E x k 1, k 1, 2 (MUST) The far user (UE2) decodes x 2 directly from received y 2. y h xw hx ( hxw) The near user (UE1) decodes x 1 from received y 1 by successively detect and subtract x 2 first. x y1 hx 1 W1 1 hx 1 2 ( hx 1 1W1) h 2 ' y 1 ' h Then decode x 1 1 from y ( hx 1 1 1W1) BS y 1 y 2 This is called successive interference cancellation (SIC) UE1 UE2 2 h 1 R1 log(1 ) N1 2 (1 ) h2 R2 log(1 ) 2 h N

37 In NOMA, In OMA, Capacity Region of NOMA 1, 1 1, NOMA with better performance than OMA 37 A. Benjebbour, Y. Saito, Y. Kishiyama, A. Li, A. Harada, and T. Nakamura, Concept and practical considerations of non orthogonal multiple access (noma) for future radio access, in 2013 International Symposium on Intelligent Signal Processing and Communications Systems (ISPACS), pp , Nov 2013.

38 Capacity Region of NOMA Symmetric channel: Capacity regions of OMA and NOMA schemes are identical. 2 2 h1 h2 N1 N 2 = 10dB Asymmetric channel: Capacity region of NOMA is larger than that of OMA. R 2 (b/s/hz) 1 h 1 N h 2 N =20dB 0dB R 1 (b/s/hz) 38

39 Code Domain Multiple Access Code domain multiple access: A group of different users signals are transmitted on the same frequency and time resources Each user s signals are spread by a sequence which can facilitate robust successive interference cancellation The spreading sequence have different properties: <RSMA> combination of low rate channel codes and scrambling codes with good correlation properties to separate different users signals (long/pseudo random sequence) <MUSA> uses low cross correlation and non binary sequences (short/fixed pattern sequence) 39

40 Sparse Code Multiple Access (SCMA) Sparse Code Multiple Access (SCMA) is a multi dimensional codebook based nonorthogonal spreading technique. Incoming bits of each user are directly mapped to dimensional vector based on each SCMA encoder, which the number of non zero entry is sparse compared to the vector length. M. Taherzadeh, H. Nikopour, A. Bayesteh, and H. Baligh, Scma codebook design, in 2014 IEEE 80th Vehicular Technology Conference (VTC Fall), pp. 1 5, Sept

41 Sparse Code Multiple Access (SCMA) Hybrid code and power domain non orthogonal multiple access Code spreading: Spreading the m coded bits transmitted on one RE in LTE to K*m bits which are transmitted on K subcarriers There are some zeros components used to make sparseness (Power domain) SCMA is issued in 2013 by Ottawa wireless R&D Center of Huawei Layer 1 Layer 2 Layer 3 Layer 4 Layer 5 Layer OFDM tone Joint design of multi dimensional modulation & low density spreading FEC SCMA Sparse codebook mapping 41 OFDM modulator 3GPP Tdoc R

42 Sparse Code Multiple Access (SCMA) Transmitter architecture Lower number of projection points : Transmission power could be reduced Reduce the complexity of ML decoding Some constellation points collide over one RE but still separable over the other RE UE1 FEC Encoder 1 b 11 b 12 SCMA MODULATION CODEBOOK MAPPING UE2 FEC Encoder 2 b 21 b 22 SCMA MODULATION CODEBOOK MAPPING SCMA block 1 SCMA block 2 UE3 FEC Encoder 3 b 31 b 32 SCMA MODULATION CODEBOOK MAPPING UE2 UE4 UE5 UE6 FEC Encoder 4 FEC Encoder 5 FEC Encoder 6 b 41 b 42 b 51 b 52 b 61 b 62 SCMA MODULATION CODEBOOK MAPPING SCMA MODULATION CODEBOOK MAPPING SCMA MODULATION CODEBOOK MAPPING UE4 UE6 OFDM tone f 42

43 Sparse Code Multiple Access (SCMA) Transmitter of SCMA S. Zhang, X. Xu, L. Lu, Y. Wu, G. He, and Y. Chen, Sparse code multiple access: An energy efficient uplink approach for 5g wireless systems, in 2014 IEEE Global Communications Conference (GLOBECOM), 43 pp , Dec 2014.

44 Sparse Code Multiple Access (SCMA) Receiver of SCMA Replace the user channel equalization and QAM demapper as SCMA demodulator Jointly detects the superposed data layers and output separate LLR results to turbo decoders of each layer Low density spreading(lds) is used LDS can reduce the complexity of advanced symbol level detectors such as message passing alogrithm (MPA), codeword level SIC somewhat reduces the need of advanced symbol level detectors RE1 RE2 Y1 Y X X

45 Sparse Code Multiple Access (SCMA) Receiver of SCMA Low density spreading(lds) is used LDS can reduce the complexity of advanced symbol level detectors such as message passing algorithm (MPA), codeword level SIC somewhat reduces the need of advanced symbol level detectors layer node resource nodes 45 F

46 Summary The new RAT in 5G shall be inherently forward compatible. The new waveforms in new RAT Multiplexed with OFDM based signaling and function as the spectrum shaping of legacy OFDMA transmission. Facilitates heterogeneous LTE downlink spectrum access with high spectrum efficiency. Non orthogonal multiple access Either power domain or code domain to improve multiuser rate region and hence spectrum efficiency. 46