Kauai, USA December 07, 2000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

3GPP2-C50-20001206-____

Title: Abstract:

Effects of Scheduling on Transmit Diversity Performance in 1xEV-DV. This contribution is presented as supporting information to Lucents 1xEVDV proposal. Achilles Kogiantis, Niranjan Joshi, and Oguz Sunay Lucent Technologies 973 386-4399 [achilles, nsjoshi, sunay] @lucent.com

Source:

Date: Recommendation:

December 07, 2000 Review and Discuss

Notice

2000 Lucent Technologies. All rights reserved. The information contained in this contribution is provided for the sole purpose of promoting discussion within the 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 TIA or 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

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

INTRODUCTION

The submitted proposals for the 1xEVDV standard give an important role to a fast packetswitching downlink channel with scheduling among the users queued packets. The scheduling operation interacts with the physical layer processes, unlike in the majority of the previously studied, and implemented, wireless systems that were designed for optimizing multiple access circuit-switched links. This interaction of the physical with the Medium Access Control (MAC), and higher, layers is evident in the case of scheduling since the scheduler operation is heavily dependent on feedback information from the physical layer. Particularly, the interaction of the scheduling process with that of the physical layer transmit diversity schemes is analyzed. Transmit diversity has been introduced in the physical layer to improve the downlink per-user performance and consequently to improve the overall capacity of the system. In the later sections, it will be shown that the combined performance of scheduling and transmit diversity does not lead to the same conclusions as a physical layer only study of the transmit diversity. A similar study leading to the same conclusions was conducted in [3]. 2 TRANSMIT DIVERSITY TECHNIQUES FOR DOWNLINK TRANSMISSION

17 18 19 20 21 22 23 24 25 26

For the purpose of this study, two transmit antennas are considered at the base station for downlink transmission for the forward packet data channel in Lucents 1xEV-DV proposal. Also, two transmit diversity techniques for the forward link are considered, namely, SpaceTime-Spreading (STS) and Selection Transmit Diversity (STD). The no diversity (single Tx antenna) transmission is considered also as a reference case. The signals that may be transmitted from the two antennas experience multipath fading and propagation losses by the scattering environment. The cumulative vector channel effects seen at the mobile's receiver, can be lumped into two variables that describe the two complex channel responses, one for each transmit antenna, namely h1 and h2 . Space-Time-Spreading, [2], is an open loop scheme and guarantees second-order diversity for each transmitted symbol. In this scenario, the mobile receiver observes a postcombining channel gain h
2

27 28

29

for each symbol equal to: h

1 2 h1 + h2 2

), which is the
2 2

30 31 32 33 34 35 36

average of the channel gains of the two paths. Selection Transmit Diversity is a closed loop scheme with binary feedback sent by the mobile. In this case, assuming error-free feedback with no delays, the observed channel gain at the mobile receiver is the better of the two paths: h
2

= max h1 , h2

). The

feedback bit and the consequent antenna selection is performed every slot. In the following, the assumption of independent channel responses from each transmit antenna is made.

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

SCHEDULING ALGORITHMS

In conjuction with the trasnmit diversity schemes that were defined previously, three different scheduling algorithms for downlink transmission are considered for the analysis: Maximum C/I Scheduler (max C/I). This scheduler essentially ranks all the users according to their instantaneous carrier-to-interference (C/I) ratios. This scheduler is optimal in obtaining the maximum network throughput [1]. Proportional Fair Scheduler (PF). This algorithm is described in detail in [4]. In short, the scheduler computes the ratio of the current supportable rate Current ratei(t), for each user i to the average throughput each user has received so far Ri(t). A user with the largest ratio gets priority in receiving data. The average throughput is updated as, Ri(t+1)=(1-1/tc)Ri(t) + transmission ratei, where tc is a constant. A user that does not receive service has 0 for his transmission ratei. Random Scheduler. As the name suggests, users are picked randomly. Essentially the performance of this scheduler is equivalent to the Round Robin scheduler that offers no advantage to users with favorable channel conditions

16 17 18 19 20 21

SIMULATION RESULTS

The Forward Packet Data Channel of Lucents 1xEVDV proposal is used as a platform for our study. The assumption of data only loading (no voice users present) is made. The link level performance is abstracted in the form of the mapping between the data rate and C/I values for an AWGN channel as shown in Table 1, and [5]. A users' initial C/I is drawn from the distribution shown in Fig. 1. Data Rate (Kbps) 38.4 76.8 102.6 153.6 204.8 307.2 614.4 921.6 1228.8 1843.2 2457.6 C/I (dB) -12.5 -9.5 -8.5 -6.5 -5.7 -4.0 -1.0 1.3 3.0 7.2 9.5

22

Table 1. Data Rate to C/I Mapping This distribution is the result of a multicell simulation, with lognormal shadowing, pathloss and antenna pattern gains. A correlated Rayleigh Jakes fading process is generated for

23 24

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

each diversity path around the initial C/I. At the system level, a simplifying assumption of single slot transmission (1.25ms) for all data rates, as opposed to the rate-dependent varying slot transmission, is made. The base station is assumed to have all user queues always full so as to transmit data continuously. Thus, no traffic model assumption is made. For a different set of number of users, fading rates, and scheduling schemes, multiple trials are performed. Each trial runs for 13 seconds.
C/I Distribution in a Multi-Cell Environment 1 0.9 0.8 0.7 0.6 CDF 0.5 0.4 0.3 0.2 0.1 0 -15

-10

-5

10

15

20

25

7 8 9

Mobile Station Received C/I (dB)

Figure 1. Cumulative Distribution of downlink received C/I based on a 19-cell scenario with lognormal shadowning and pathloss modeling. 4.1 Results The performance of various transmit diversity schemes in conjunction with the various scheduling algorithms is measured in terms of the mean sector/network throughput. A users received signal strength distribution, which includes shadowing and fading, will contribute significantly to the sector throughput. This received C/I distribution, is also influenced by the employed transmit diversity scheme. This is evident from the pdf curves representing random scheduler (i.e. no scheduler) in Fig. 2. The STD scheme clearly displays the highest mean C/I. The STS, as expected, shows a smaller variance and has a higher mean compared to the single antenna case. In Fig. 2 it is shown how the scheduler utilizes this C/I information. It is evident from Figs. 3 and 4 that, for the random and max C/I scheduler the performance of STD is superior to the other schemes. Moreover, the STS performs slightly better that the single antenna case. For the PF scheduler, which accounts for more than just C/I information, the single antenna performance is superior to the STS scheme. From all the figures, it is evident that an efficient scheduler exploits more efficiently the tails of the per user C/I distribution, as shaped with the use of STD. This is not the case though, for the STS scheme that suppresses the distribution tails, while increasing its mean. Fig. 3 shows that for all schemes the mean sector throughput increases with increased cell loading. Moreover the throughput curve saturates quickly either due to the assumption of rate set limitation to 2457.6 kbps as well as the scheduler behavior. The throughput

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

remains insensitive to the increase of fading rates, as shown in Fig. 4. Only the PF scheduler shows a measurable reduction in the network throughput for low to medium fading rates. For all sector/network loadings and all fading rates, STD is uniformly superior in throughput achieved, by at least 5% versus the single transmit antenna case.
Scheduler output C/I distribution. Fading rate=3Hz, 16 users 0.07 1Tx max C/I STS max C/I STD max C/I 1Tx Random 0.05 STS Random STD Random 1Tx PF 0.04 STS PF STD PF

0.06

PDF
0.03 0.02 0.01 0 -20
5 6 7

-15

-10

-5

10

15

20

25

Scheduler Output C/I (dB)

Figure 2. Scheduler Output distribution for various cases of transmit diversity and scheduling algorithms.
Mean Throughput vs. Cell Loading for 3Hz fading rate 2600 2400 2200 2000

1Tx max C/I STS max C/I 1Tx Random STS Random STD max C/I STD Random 1Tx PF STS PF STD PF

Mean Throughput (Kbps)

1800 1600 1400 1200 1000 800 600 10 20 30 40

50

60

70

8 9 10

Number of users

Figure 3. Combined transmit diversity and scheduling performance as a function of the number of active users for fading rate of 3Hz.

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

Mean Throughput (Kbps)

1Tx max C/I S TS max C/I 1Tx Random S TS Random S TD max C/I S TD Random 1Tx PF S TS PF S TD PF

10

20

30

40 50 60 Fading Rate (Hz)

70

80

90

100

Figure 4. Combined transmit diversity and scheduling performance as a function of the fading rate for a fixed sector loading of 16 active users. 5 CONCLUSIONS

4 5 6 7 8 9 10 11 12 13 14 15 16 17

A study on the interaction of three scheduling algorithms with three different transmission configurations (with and without transmit diversity) on a data only case for the 1xEVDV system was conducted. The simulation results indicate that without transmit diversity (i.e. only one transmit antenna) the resulting network throughput is equivalent to the one with the most efficient open loop two-branch transmit diversity scheme, STS. On the other hand, the closed loop STD is shown to offer distinct advantages in conjunction with the scheduling operation. For low to medium cell loading, the network throughput can be 510% higher when STD is deployed. The performance differences among the diversity schemes are dependent on the resulting per-user C/I distribution profiles. The greedy schedulers take advantage of the high-value tails of the per-user C/I distributions while STS attempts to shrink these tails. Also, it was shown that the network throughput is not affected by the Doppler rate, except for the case of the PF scheduler that is more sensitive in the low to medium speeds. 6 REFERENCES:

18 19 20 21 22 23

1. D. Tse, and S. Hanly, ``Multiaccess Fading Channels-Part I: Polymatroid Structure, Optimal Resource Allocation and Throughput Capacities,'' IEEE Transactions on Information Theory, Vol. 44, No. 7, November 1998. 2. Lucent Technologies Inc., ``Down Link Improvement through Space Time Spreading,'' Standards Contribution 3GPP2-C30-19990817-014.

Kauai, USA December 07, 2000

3GPP2-C50-20001206-____

3. H. Huang, H. Viswanathan, A. Blanksby, and M. Haleem, ``Multiple Antenna Enhancements to High Rate Packet Data CDMA System,'' submitted to the Journal of VLSI Signal Processing, September 2000. 4. A. Jalali, R. Padovani, and R. Pankaj, ``Data Throughput of CDMA-HDR: A High Efficiency-High Data Rate Personal Communication Wireless System,'' IEEE Vehicular Technology Conference, Tokyo, Japan, May 2000. 5. P. Bender, P. Black, M. Grob, and R. Padovani, ``CDMA/HDR: A Bandwith-Efficient High-Speed Wireless Data Service for Nomadic Users,'' IEEE Communications Magazine, July 2000. 6. N. Joshi, S. Kadaba, S. Patel, and G. Sundaram, Downlink Scheduling in CDMA Data Networks, Mobicom 2000 Conference, Boston, MA, August 2000.