Physical Level Performance Analysis of Satellite High Speed Downlink Packet Access (S-HSDPA)

Transcription

1 Physical Level Performance Analysis of Satellite High Speed Downlink Packet Access (S-HSDPA) A. Azizan Supervised by A. Quddus and B. Evans Centre for Communication Systems University of Surrey Guildford, GU2 7XH, Surrey, UK Mobile CCSR

2 Outline of Presentation Introduction (Sat Comms & HSDPA) HSDPA via Satellites Link-Level Simulation Set up Contribution of Work Link Level Performance Low Complexity Advanced Receivers Orthogonality Factor 27/11/08 2 Mobile

3 Satellite The global mobile personal communication via satellite (GMPCS) Weaknesses in capturing the market Successful in niche geographical areas and broadcasting Thuraya satellites Geo Mobile Radio-1 (GMR-1) from Global System for Mobile (GSM) Geo Mobile Packet Radio Services (GMPRS) from General Packet Radio Service (GPRS) Several European Union (EU) projects convergence of mobile satellite communication with terrestrial mobile communication networks SATIN & MODIS project developed Satellite-UMTS (SUMTS) system with Intermediate module repeaters (IMRs) to penetrate the urban areas The satellite component of the IMT-2000 standardization framework Developing SUMTS standards Keeping close commonality with their terrestrial counterparts SDMB trials Conducted with DVB-SH receivers 2007/8 EUTELSAT W2A satellite launch and commercial services expected in 27/11/08 3 Mobile 2009/10

4 HSDPA HSDPA is a new radio bearer specified for 3GPP T-UMTS Provides higher capacity, higher data rates, low delay Peak data rates of Mb/s Average data rates of 2 3 Mb/s Supports mainly 2 types of services: Streaming Background services (e.g.internet browsing and large size high speed file transfer) Main Elements of HSDPA Higher Order Modulation Shorter TTI = 2 ms 2 ms Fast ARQ with soft combining Link Adaptation 27/11/08 4 Mobile

5 Scope of This Study Consider the applicability of HSDPA To increase the capacity for high data rate applications over the S- UMTS / S-DMB downlink air-interface, whilst other project looked at OFDM Envisions the usage of IMR (terrestrial repeaters) Study carried out as part of EC FP-6 project MAESTRO System Architecture Satellite Unicast + Multicast traffic A ir interface IMR 1 IMR 2 IMR 3 Gateway Intermediate Module Repeater T-UMTS Node B IMR Ref IMR 5 IMR 3 UMTS core network T erminal IMR 4 27/11/08 5 Mobile

6 HSDPA via Satellites - Issues Power, db Chip Delay in Chip Resolution GEO round trip delay is large ~ 0.5 sec Cannot track small scale multipath fading link adaptation not possible. Causes the number of parallel ARQ processes to be large resulting in large memory requirements for a UE The IMRs cause significant multipath fading resulting higher intracell interference at receiver compared to terrestrial channels Maestro Case 4 channel (Sat + 3 IMRs), urban street canyon environment with rich multipath and a speed of 30 km/h is used. Mobile satellite communication system is power limited Transmit power utilization need to be managed carefully in order to satisfy the QoS of multiple users Two main study parameters (Throughput and Average Air Interface Delay against E c / I or ) 27/11/08 6 Mobile

7 Simulation Block Diagram Link level simulations are carried out by following the 3GPP specifications very closely. 27/11/08 7 Mobile

8 Diversity and Multicodes Receive antenna diversity Implemented at UE Performance depends on the correlation values between the fading signals at the receive antennas Space Time Transmit Diversity (STTD) A requirement in WCDMA implemented using two GEO satellite, each emulating one antenna Multiple physical channel transmission Within a TTI interval (5, 10, and 15) Maintaining the transmitted data size (decreases the coding rate - Turbo coding gain) Antenna 1 Antenna 2 27/11/08 8 Mobile d Inf. Bit Payload Receiver S 1 RV Selection Physical Channel Segmentation Satellite 1 S 2 Satellite multicodes r 2 r 4 r 1 S 1 S 2 TX Diversity Encoder - S 2 * S 1 * 960 Scatterer 1 r 3 Scatterer 2 Terminal Antenna multicodes r s 10 multicodes Transmitter TX Diversity De coder Coding Rate = 2/3 Coding Rate = 1/3 Coding Rate = 2/3

9 Number of Retransmissions For QPSK, more than 2 retxns. do not provide much improvement in throughput whereas for 16-QAM, performance is improved up to 3 rd retxn. QPSK 16-QAM 27/11/08 9 Mobile

10 Receive Antenna Diversity 16-QAM outperforms QPSK only at very high values of E c / I or Throughput reduces as the antenna correlation increases QPSK 16-QAM 27/11/08 10 Mobile

11 Receive Antenna Diversity (cont.) Average air interface delay is higher for 16QAM even though its throughput is higher with respect to QPSK QPSK 16-QAM 27/11/08 11 Mobile

12 Space Time Transmit Diversity (STTD) STTD not beneficial in multipath propagation environment and throughput gain obtained only in Ricean LOS channel 27/11/08 12 Mobile

13 Multicodes Transmission Increasing multicodes while maintain TBS has increase Turbo coding protection and increased temporal diversity gaining around 1.5 db 10MC with larger TBS=6438 better at high power allocation The increase in multicodes for same TBS=3202 lowers the average air interface excess delay (reaching minimal value of 2 ms). 27/11/08 13 Mobile

14 Summary: Link Level Performance QPSK facilitates higher throughput than 16-QAM (more robust in the dispersive multipath channel) Receive antenna diversity at the UE provides substantial performance improvements meanwhile STTD does not increase the throughput in multipath channel environments The increase in number of multicodes increases the throughput due to the repetition of the symbols, but at the expense other user/physical channels Lower average air interface excess delay achieved by increasing the multicodes and receive antenna diversity No HARQ/retransmission Possible to utilize the S-HSDPA for Satellite Digital Multicast Broadcast (S- DMB) systems Achieve a higher average throughput compared to the current FACH results (384 kbps) 27/11/08 14 Mobile

15 Receiver Architecture- Issues Highly dispersive IMR channel Non-zero cross correlation between synchronously transmitted multi-user signals in the downlink The intracell interference severely decreases the S-HSDPA throughput Low complexity advanced receivers (Multipath interference cancellers and chip level adaptive linear equalizers) Retains the orthogonality of the received signal impaired by the IMR channel Improve the S-HSDPA average user throughput Linear equalizer prefered as non-linear equalizers (decision feedback equalizer) rely on a-priori knowledge of the constellation of the desired reference signal LMMSE equalizer computationaly expensive and CMA equalizer not possible for WCDMA systems 27/11/08 15 Mobile

16 RAKE Matched Filter Manipulation of multipath diversity Maximize the received signal to background (thermal) noise ratio at the receiver (matched filtering) Intracell interference cannot be discriminated and suppressed. Low orthogonality between the physical channels r Correlator Channel Comp. Delay Equaliser Σ I I Spreading Code Channel Estimator Σ Q Q Finger 1 Combiner Finger 2 Finger 3 Delay Estimator 27/11/08 16 Mobile

17 CPICH NLMS Chip-Level Adaptive Equalizer r Equalizer z Correlator y N LMS w(n+1) - s Pilot Signal Spreading Sequence w ( n + 1) = w ( n ) + µ r ( n ) e ( n ) r H ( n ) r ( n ) Chip level CPICH (typically 10% of the total transmitted power) used as training data Does not need channel estimation 27/11/08 17 Mobile

18 Griffith s Chip-Level Adaptive Equalizer r Equalizer ( )* z Correlator y w Adaptation Process - p Channel Estimator Channel Coefficient w ( n + 1) = w ( n) µ ( z * ( n) r ( n) p ) Low complexity calculation similar to the LMS algorithm Blind equalizer (does not require training sequence) but requires the channel estimation values (could be obtained via CPICH) 27/11/08 18 Mobile

19 r Multipath Interference Canceller Delay Delay To next stage CEIGU CEIGU ˆ ( I 1 ) l = 1, 2,..., L ˆ ( 2 ) I l = 1, 2,..., L 1 st Stage l i L ( ) I ˆ ( 1 ) l l = 1 α (1) 2 nd Stage r ( p + 1) l Similar to parallel interference cancellation (PIC), Consists of several CEIGUs (Channel estimation and interference replica generation units) Hard-decisions used to estimate the symbols for interference replica generation Keep the complexity of MPIC comparable to equalizers 27/11/08 19 Mobile ( p ) ( p ) ( t ) = r ( t ) α Iˆ ( t ˆ τ L j = 1 j l j j )

20 RAKE vs Advanced Receivers Superiority of chip-level adaptive equalizers compared to MPIC Griffith s equalizer performs the best MPIC suppresses the multipath interference from intended physical channel only MPIC exceeds the performance of the Griffith s equalizer at high E c / Ior 27/11/08 20 Mobile

21 Different CPICH Power Allocation 20% power allocation gives much better performance than 10% power allocation, whereas 5% is giving worst performance than RAKE. 27/11/08 21 Mobile

22 Different Channel Estimation for Griffith s Equalizer Griffith s equalizer maintains superior performance with CPICH channel estimation Griffiths Perfect Griffiths CPICH Estimation Throughput, kbps Ec/Ior, db 27/11/08 22 Mobile

23 Complexity Calculations Superiority of the Griffiths equalizer (achieves a high throughput and low average interface delay with relatively low computational complexity increase compared to other advanced receivers) Total per iteration Increase (compared to RAKE) Receiver x + / x + / RAKE receiver CPICH NLMS equalizer x40 x48 +1 Griffiths equalizer x31 x38 MPIC x35 x40 Note: I = 6 (total number of RAKE fingers) N = 53 (total filter length) L = 8 (total channel tap length) P = 2 (total MPIC stages) Q = 5 (total HS-DSCH physical channel) 27/11/08 23 Mobile

24 Summary: Advanced Receivers Equalization based receivers perform better in comparison to the interference cancellation and multipath diversity receivers. The CPICH NLMS equalizer is better than RAKE receiver for the S-UMTS system and an increase of CPICH power can further increase the throughput performance Griffiths equalizer provides the best trade-off in performance (up to 2.5 db over the RAKE) and complexity (up to 31x (x) and 38x (+)) and maintains performance with the standard CPICH channel estimation MPIC performs marginally better than RAKE receiver due to hard decision errors but performs better that the equalizers at high E c / Ior 27/11/08 24 Mobile

25 Orthogonality Factor Measure of loss of spreading codes orthogonality due to the intracell interference power 0 means a totally orthogonal or no downlink intracell interference Used for budget calculations and downlink capacity estimation The highly dispersive nature of the IMR channel Distorts the orthogonality of signals Power limitation of GEO satellites requires accurate downlink capacity calculations Various methods of obtaining OF value Pedersen, obtained at system level - utilizing the generic power delay profile Passerini used amplitude power delay profile of the multipath channel Metha used an analytical expression and link level simulation Different receiver architectures (receive antenna diversity and channel equalizers) have a considerable impact on the OF statistics not been 2 dealt with N Instantaneous orthogonality factor [METH03] = M ( P P ) S F w * α 27/11/08 25 Mobile β ( i) o tot P i i ~ I or ( i) n = 1 N ~ F ξ i n n = 1 n w * n 2

26 Orthogonality Factor S-HSDPA Pedestrian B (largest delay spread in the 3GPP terrestrial channels) gives lower OF mean and median values in contrast to the IMR Case 4 channel Increasing the multicodes to 10 reduces OF statistics (less intracell interference) RAKE receiver Mean Median Standard Deviation IMR Case Pedestrian B HS-DSCH, IMR Case /11/08 26 Mobile

27 OF for S-HSDPA (Receive Ant Div) OF statistics decreases for two receive antennas diversity compared to single receive antenna 66% decrease for mean 57% decrease for median 90% of the OF values Below 0.4 (two receive antennas) Below 1.2 (single antenna) RAKE receiver Mean Median Standard Deviation Single receive antenna Two receive antenna /11/08 27 Mobile

28 OF for S-HSDPA (Equalizers) Equalizer restore orthogonality of multipath faded signal Distribution of the low OF values are significantly larger than RAKE The mean of the equalizers is quite high compared the median values High variance of the OF values Mean OF inaccurate Equalizer adaptation to track the channel Mean Median Standard Deviation RAKE receiver CPICH NLMS Equalizer Griffiths Equalizer /11/08 28 Mobile

29 Summary: Orthogonality Factor The OF statistics for the S-HSDPA system is generally higher than terrestrial HSDPA More dispersive channel encountered in the IMR propagation environment The increase in the number of HS-DSCH physical channels effectively improves the OF statistics The OF statistics and time-variant distribution improves significantly for receive antenna diversity. The equalizer based receivers have lower OF statistics and values to the RAKE receiver but the mean OF is higher due to the nature of the channel adaptation algorithm. 27/11/08 29 Mobile

30 List of Publications 1. Azizan A, Quddus A, Evans B, Link Level Performance Analysis of Satellite High Speed Downlink Packet Access (S_HSDPA), 24th AIAA International Satellite Systems Conference (ICSSC), San Diego, California, June Azizan A, Quddus A, Evans B, Chip Level Adaptive Equalization for Satellite High Speed Downlink Packet Access (S_HSDPA), 25th AIAA International Satellite Systems Conference (ICSSC), Seoul, South Korea, Apr Azizan A, Quddus A, Evans B, Multipath Interference Canceller and Chip Level Equalizer for Satellite High Speed Downlink Packet Access (S_HSDPA), 26th AIAA International Satellite Systems Conference (ICSSC), San Diego, California, June Azizan A, Quddus A, Evans B, Satellite High Speed Downlink Packet Access Physical Layer Performance Analysis, 4th Advanced Satellite Mobile Systems Conference, Bologna, Italy, August /11/08 30 Mobile

31 Question Time 27/11/08 31 Mobile

32 Summary of work on S-HSDPA Basic link level performance by considering HARQ, QPSK/16QAM Receive antenna diversity Transmit antenna diversity (STTD) Multicodes Low Complexity Advanced Receivers Chip-level adaptive equalizer (CPICH NLMS equalizer and Griffiths equalizer) Multipath Interference Canceller Orthogonality Factor for S-HSDPA cases Receive antenna diversity Multicodes Chip-level adaptive equalizer 27/11/08 32 Mobile