Lucent1xDiversityScheduling

download Lucent1xDiversityScheduling

of 7

Transcript of Lucent1xDiversityScheduling

  • 8/6/2019 Lucent1xDiversityScheduling

    1/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    1

    1

    2

    3

    4

    Title: Effects of Scheduling on Transmit Diversity Performance in 1xEV-DV.5

    6

    Abstract: This contribution is presented as supporting information to Lucents71xEVDV proposal.8

    9

    Source: Achilles Kogiantis, Niranjan Joshi, and Oguz Sunay10

    Lucent Technologies11973 386-439912[achilles, nsjoshi, sunay] @lucent.com13

    14

    Date: December 07, 200015

    16

    Recommendation: Review and Discuss17

    Notice

    2000 Lucent Technologies. All rights reserved.

    The information contained in this contribution is provided for the sole purpose of promoting discussion withinthe 3GPP2 and its Organization Partners and is not binding on the contributor. The contributor reserves the

    right to add to, amend, or withdraw the statements contained herein.

    The contributor grants a free, irrevocable license to 3GPP2 and its Organization Partners to incorporate text or

    other copyrightable material contained in the contribution and any modifications thereof in the creation of TIAor 3GPP2 publications; to copyright and sell in Organizational Partners name any Organizational Partners

    standards publication even though it may include portions of the contribution; and at the Organization Partners

    sole discretion to permit others to reproduce in whole or in part such contributions or the resulting

    Organizational Partners standards publication

  • 8/6/2019 Lucent1xDiversityScheduling

    2/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    2

    1

    1 INTRODUCTION2The submitted proposals for the 1xEVDV standard give an important role to a fast packet-3

    switching downlink channel with scheduling among the users queued packets. The4

    scheduling operation interacts with the physical layer processes, unlike in the majority of5

    the previously studied, and implemented, wireless systems that were designed for6

    optimizing multiple access circuit-switched links. This interaction of the physical with the7

    Medium Access Control (MAC), and higher, layers is evident in the case of scheduling since8

    the scheduler operation is heavily dependent on feedback information from the physical9

    layer. Particularly, the interaction of the scheduling process with that of the physical layer10

    transmit diversity schemes is analyzed. Transmit diversity has been introduced in the11

    physical layer to improve the downlink per-user performance and consequently to improve12

    the overall capacity of the system. In the later sections, it will be shown that the combined13

    performance of scheduling and transmit diversity does not lead to the same conclusions as14

    a physical layer only study of the transmit diversity. A similar study leading to the same15

    conclusions was conducted in [3].16

    2 TRANSMIT DIVERSITY TECHNIQUES FOR DOWNLINK TRANSMISSION17For the purpose of this study, two transmit antennas are considered at the base station for18

    downlink transmission for the forward packet data channel in Lucents 1xEV-DV proposal.19

    Also, two transmit diversity techniques for the forward link are considered, namely, Space-20

    Time-Spreading (STS) and Selection Transmit Diversity (STD). The no diversity (single Tx21

    antenna) transmission is considered also as a reference case. The signals that may be22

    transmitted from the two antennas experience multipath fading and propagation losses by23

    the scattering environment. The cumulative vector channel effects seen at the mobile's24

    receiver, can be lumped into two variables that describe the two complex channel25

    responses, one for each transmit antenna, namely 1h and 2h .26

    Space-Time-Spreading, [2], is an open loop scheme and guarantees second-order diversity27

    for each transmitted symbol. In this scenario, the mobile receiver observes a post-28

    combining channel gain2

    h for each symbol equal to: ( )222

    1

    2

    2

    1hhh += , which is the29

    average of the channel gains of the two paths.30

    Selection Transmit Diversity is a closed loop scheme with binary feedback sent by the31

    mobile. In this case, assuming error-free feedback with no delays, the observed channel32

    gain at the mobile receiver is the better of the two paths: (2

    2

    2

    1

    2

    ,max hhh = . The33

    feedback bit and the consequent antenna selection is performed every slot.34

    In the following, the assumption of independent channel responses from each transmit35

    antenna is made.36

  • 8/6/2019 Lucent1xDiversityScheduling

    3/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    3

    3 SCHEDULING ALGORITHMS1In conjuction with the trasnmit diversity schemes that were defined previously, three2

    different scheduling algorithms for downlink transmission are considered for the analysis:3

    Maximum C/I Scheduler (max C/I). This scheduler essentially ranks all the users4

    according to their instantaneous carrier-to-interference (C/I) ratios. This scheduler5

    is optimal in obtaining the maximum network throughput [1].6

    Proportional Fair Scheduler (PF). This algorithm is described in detail in [4]. In7

    short, the scheduler computes the ratio of the current supportable rate Current8

    ratei(t), for each user i to the average throughput each user has received so far Ri(t).9

    A user with the largest ratio gets priority in receiving data. The average throughput10

    is updated as, Ri(t+1)=(1-1/tc)Ri(t) + transmission ratei, where tc is a constant. A user11

    that does not receive service has 0 for his transmission ratei.12

    Random Scheduler. As the name suggests, users are picked randomly. Essentially13

    the performance of this scheduler is equivalent to the Round Robin scheduler that14

    offers no advantage to users with favorable channel conditions15

    4 SIMULATION RESULTS16The Forward Packet Data Channel of Lucents 1xEVDV proposal is used as a platform for17

    our study. The assumption of data only loading (no voice users present) is made. The link18

    level performance is abstracted in the form of the mapping between the data rate and C/I19

    values for an AWGN channel as shown in Table 1, and [5]. A users' initial C/I is drawn from20

    the distribution shown in Fig. 1.21

    Data Rate (Kbps) C/I (dB)

    38.4 -12.5

    76.8 -9.5

    102.6 -8.5

    153.6 -6.5

    204.8 -5.7

    307.2 -4.0

    614.4 -1.0

    921.6 1.3

    1228.8 3.0

    1843.2 7.2

    2457.6 9.5

    Table 1. Data Rate to C/I Mapping22

    This distribution is the result of a multicell simulation, with lognormal shadowing, pathloss23

    and antenna pattern gains. A correlated Rayleigh Jakes fading process is generated for24

  • 8/6/2019 Lucent1xDiversityScheduling

    4/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    4

    each diversity path around the initial C/I. At the system level, a simplifying assumption of1

    single slot transmission (1.25ms) for all data rates, as opposed to the rate-dependent2

    varying slot transmission, is made. The base station is assumed to have all user queues3

    always full so as to transmit data continuously. Thus, no traffic model assumption is made.4

    For a different set of number of users, fading rates, and scheduling schemes, multiple trials5

    are performed. Each trial runs for 13 seconds.6

    -15 -10 -5 0 5 10 15 20 250

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    1

    Mobile Station Received C/I (dB)

    CDF

    C/I Distribution in a Multi-Cell Environment

    7

    Figure 1. Cumulative Distribution of downlink received C/I based on a 19-cell8

    scenario with lognormal shadowning and pathloss modeling.9

    4.1 Results10 The performance of various transmit diversity schemes in conjunction with the various11

    scheduling algorithms is measured in terms of the mean sector/network throughput. A12users received signal strength distribution, which includes shadowing and fading, will13

    contribute significantly to the sector throughput. This received C/I distribution, is also14

    influenced by the employed transmit diversity scheme. This is evident from the pdf curves15

    representing random scheduler (i.e. no scheduler) in Fig. 2. The STD scheme clearly16

    displays the highest mean C/I. The STS, as expected, shows a smaller variance and has a17

    higher mean compared to the single antenna case. In Fig. 2 it is shown how the scheduler18

    utilizes this C/I information. It is evident from Figs. 3 and 4 that, for the random and max19

    C/I scheduler the performance of STD is superior to the other schemes. Moreover, the STS20

    performs slightly better that the single antenna case. For the PF scheduler, which accounts21

    for more than just C/I information, the single antenna performance is superior to the STS22

    scheme. From all the figures, it is evident that an efficient scheduler exploits more23

    efficiently the tails of the per user C/I distribution, as shaped with the use of STD. This is24

    not the case though, for the STS scheme that suppresses the distribution tails, while25

    increasing its mean.26

    Fig. 3 shows that for all schemes the mean sector throughput increases with increased cell27

    loading. Moreover the throughput curve saturates quickly either due to the assumption of28

    rate set limitation to 2457.6 kbps as well as the scheduler behavior. The throughput29

  • 8/6/2019 Lucent1xDiversityScheduling

    5/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    5

    remains insensitive to the increase of fading rates, as shown in Fig. 4. Only the PF1

    scheduler shows a measurable reduction in the network throughput for low to medium2

    fading rates. For all sector/network loadings and all fading rates, STD is uniformly superior3

    in throughput achieved, by at least 5% versus the single transmit antenna case.4

    -20 -15 -10 -5 0 5 10 15 20 250

    0.01

    0.02

    0.03

    0.04

    0.05

    0.06

    0.07

    Scheduler Output C/I (dB)

    PDF

    Scheduler output C/I distr ibution. Fading rate=3Hz, 16 users

    1Tx max C/I

    STS max C/ I

    STD max C/I

    1Tx Random

    STS Random

    STD Random

    1Tx PF

    STS PF

    STD PF

    5

    Figure 2. Scheduler Output distribution for various cases of transmit diversity and6

    scheduling algorithms.7

    10 20 30 40 50 60 7060 0

    80 0

    1000

    1200

    1400

    1600

    1800

    2000

    2200

    2400

    2600

    Number of users

    MeanThroughput(Kbps)

    Mean Throughput vs. Cell Loading for 3Hz fading rate

    1Tx max C/I

    STS max C/I

    1Tx Random

    STS Random

    STD max C/I

    STD Random

    1Tx PF

    STS PF

    STD PF

    8

    Figure 3. Combined transmit diversity and scheduling performance as a function of9

    the number of active users for fading rate of 3Hz.10

  • 8/6/2019 Lucent1xDiversityScheduling

    6/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    6

    0 10 20 30 40 50 60 70 80 90 100

    800

    1000

    1200

    1400

    1600

    1800

    2000

    2200

    2400

    2600

    Fading Rate (Hz)

    MeanThroughput(Kbps)

    Mean Throughput vs. fading rate for 16 users

    1Tx max C/ISTS max C/I

    1Tx Random

    STS Random

    STD max C/I

    STD Random

    1Tx PF

    STS PF

    STD PF

    1

    Figure 4. Combined transmit diversity and scheduling performance as a function of2

    the fading rate for a fixed sector loading of 16 active users.3

    5 CONCLUSIONS4A study on the interaction of three scheduling algorithms with three different transmission5

    configurations (with and without transmit diversity) on a data only case for the 1xEVDV6

    system was conducted. The simulation results indicate that without transmit diversity (i.e.7

    only one transmit antenna) the resulting network throughput is equivalent to the one with8

    the most efficient open loop two-branch transmit diversity scheme, STS. On the other hand,9the closed loop STD is shown to offer distinct advantages in conjunction with the10

    scheduling operation. For low to medium cell loading, the network throughput can be 5-11

    10% higher when STD is deployed. The performance differences among the diversity12

    schemes are dependent on the resulting per-user C/I distribution profiles. The greedy13

    schedulers take advantage of the high-value tails of the per-user C/I distributions while14

    STS attempts to shrink these tails. Also, it was shown that the network throughput is not15

    affected by the Doppler rate, except for the case of the PF scheduler that is more sensitive16

    in the low to medium speeds.17

    6 REFERENCES:181. D. Tse, and S. Hanly, ``Multiaccess Fading Channels-Part I: Polymatroid Structure,19

    Optimal Resource Allocation and Throughput Capacities,'' IEEE Transactions on20

    Information Theory, Vol. 44, No. 7, November 1998.21

    2. Lucent Technologies Inc., ``Down Link Improvement through Space Time Spreading,''22

    Standards Contribution 3GPP2-C30-19990817-014.23

  • 8/6/2019 Lucent1xDiversityScheduling

    7/7

    Kauai, USA

    December 07, 2000 3GPP2-C50-20001206-____

    7

    3. H. Huang, H. Viswanathan, A. Blanksby, and M. Haleem, ``Multiple Antenna1

    Enhancements to High Rate Packet Data CDMA System,'' submitted to the Journal of2

    VLSI Signal Processing, September 2000.3

    4. A. Jalali, R. Padovani, and R. Pankaj, ``Data Throughput of CDMA-HDR: A High4

    Efficiency-High Data Rate Personal Communication Wireless System,'' IEEE Vehicular5

    Technology Conference, Tokyo, Japan, May 2000.6

    5. P. Bender, P. Black, M. Grob, and R. Padovani, ``CDMA/HDR: A Bandwith-Efficient7

    High-Speed Wireless Data Service for Nomadic Users,'' IEEE Communications Magazine,8

    July 2000.9

    6. N. Joshi, S. Kadaba, S. Patel, and G. Sundaram, Downlink Scheduling in CDMA Data10

    Networks,Mobicom 2000 Conference, Boston, MA, August 2000.11