In this section, we estimate the impact on system capacity of a WIMAX network of some of the MIMO features that are part of the IEEE 802.16e-2005 standard. See Chapter 8 for a detailed description of the various open- and closed-loop MIMO schemes.
The average per sector throughput for the basic and various enhanced configurations is shown in Figure 12.9 and Figure 12.10. Based on these results, one can conclude that both Table 12.4 Average Throughput per BS Site for Handheld and Desktop Device with Band AMC Subcarrier Permutation in DL and UL
Scenario Handheld Device Desktop Device DL (Mbps) UL (Mbps) DL (Mbps) UL (Mbps)
(1,1,3) reuse in Ped B 131.47 21.13 146.77 23.59
(1,3,3) reuse in Ped B 57.09 9.18 57.08 9.17
(1,1,3) reuse in Ped A 139.88 22.48 150.26 24.15
(1,3,3) reuse in Ped A 57.62 9.26 57.86 9.30
Table 12.5 Fifth and Tenth Percentile Data Rate for Basic Configuration Scenario Handheld Device Desktop Device
5% (Mbps) 10% (Mbps) 5% (Mbps) 10% (Mbps)
(1,1,3) reuse in Ped B 0.025 0.075 0.146 0.252
(1,3,3) reuse in Ped B 0.449 0.525 0.480 0.550
(1,1,3) reuse in Ped A 0.025 0.075 0.036 0.220
(1,3,3) reuse in Ped A 0.351 0.490 0.406 0.517
Figure 12.7 Comparison of PUSC and band AMC subcarrier permutation for handheld form factor
Figure 12.8 Comparison of proportional fairness and round-robin scheduling for handheld device
0 2 4 6 8 10 12 14 16 18 20
(1,1,3) Ped B (AMC)
(1,1,3) Ped A (AMC )
(1,1,3) Ped B (PUSC)
(1,1,3) Ped A (PUSC)
Throughput per 10MHz TDD Channel (Mbps)
Downlink Uplink
0 2 4 6 8 10 12 14 16 18 20
(1,1,3) Ped B (AMC PF)
(1,1,3) Ped A (AMC PF)
(1,1,3) Ped B (AMC RR)
(1,1,3) Ped A (AMC RR)
Throughput per 10MHz TDD Channel (Mbps)
Downlink Uplink
Figure 12.9 Downlink average throughput per sector for various MIMO configuration
Figure 12.10 Uplink average throughput per sector for various MIMO configurations
0 5 10 15 20 25 30 35 40
2 × 2 Open-Loop MIMO
2 × 4 Open-Loop MIMO
4 × 2 Open-Loop MIMO
4 × 2 Closed-Loop MIMO
Throughput per 10MHz TDD Channel (Mbps)
Ped B Multipath Channel Ped A Multipath Channel
0 1 2 3 4 5 6 7 8 9 10
1 × 2 Open-Loop MIMO
1 × 4 Open-Loop MIMO
2 × 4 Open Loop MIMO
Throughput per 10MHz TDD Channel (Mbps)
Ped B Multipath Channel Ped A Multipath Channel
receive diversity and transmit diversity improve the average throughput of a WiMAX network.
By increasing the number of transmit antennas from two to four, the per sector throughput improves by 50 percent. Similarly, by increasing the number of receive antennas from two to four, the per sector throughput is increased by 80 percent. However, based on Figure 12.11 and Figure 12.12, it is evident that for the basic 2 × 2 open-loop configuration, the fifth and tenth percentile DL data rates are not improved by increasing either transmit or receive diversity order.
Thus, one can conclude that transmit diversity with antennas in DL is not sufficient to improve the cell-edge data rate in the case of (1,1,3) reuse. Although receive diversity with four antennas somewhat improves the tenth percentile data rate, it is still not sufficient to improve the cell-edge data rate when (1,1,3) frequency reuse is implemented.
However, when closed-loop MIMO with four antennas at the BS is used, the per sector through- put and the fifth and tenth percentile data rates are significantly improved. The average per sec- tor throughput is improved by 130 percent, and the cell-edge data rate per subchannel is high enough to provide reliable broadband services.
Clearly, the 4 × 2 closed-loop MIMO feature provides significant improvement in the per sector throughput and percentile data rates, compared to both the open-loop 2 × 4 and open-loop 4 × 2 MIMO modes, because the transmitter is able to choose the optimum precoding matrix or beamforming vector, in order to increase the link throughput. In this case, we assume that a single precoding matrix or beamforming vector is chosen for each 2 × 3 band AMC subchannel.5
The DL simulation results shown in Figure 12.9–Figure 12.12 assume that feedback for the quantized MIMO channel is provided by the receiver once every frame (5msec). (See Section 8.9 for the quantized channel-feedback-based closed-loop MIMO solution). The UL enhanced configuration 3 uses a 2 × 4 open-loop MIMO. The increased performance over other enhanced profiles comes from increasing the number transmit antennas in the UL from one to two. The UL throughput results do not account for the fact that a part of the UL bandwidth is used by the closed-loop MIMO feedback.
Table 12.7 and Table 12.8 show the average throughputs per cell site and the percentile data rates for the various profiles. The biggest impact of the closed-loop MIMO appears to be on the percentile data rate (Table 12.8). Based on the system-level performance of a WiMAX network, one can conclude that a (1,1,3) frequency reuse will not be able to provide carrier-grade reliabil- ity and guaranteed data rate unless closed-loop MIMO features of IEEE 802.16e-2005 are used.
Table 12.6 Average Throughput per Sector for Band AMC with PF and RR Schedulers
Ped B Ped A
DL (Mbps) UL (Mbps) DL (Mbps) UL (Mbps)
Proportional-fairness scheduler 14.61 2.35 15.54 2.50
Round-robin scheduler 11.96 1.92 12.66 2.04
5. See Chapter 9 for a more detailed description of the band AMC subcarrier permutation mode.
Figure 12.11 User DL data rate per subchannel for enhanced profile in Ped B
Figure 12.12 User DL data rate per subchannel for enhanced profile in Ped A
0.0 0.2 0.4 0.6 0.8 1.0 1.2
0.00 0.50 1.00 1.50 2.00 2.50
User Data Rate per Subchannel (Mbps)
Cumulative Distribution Function
2 × 2 OL MIMO 2 × 4 OL MIMO 4 × 2 OL MIMO 4 × 2 CL MIMO 10th Percentile Data Rate
0.0 0.2 0.4 0.6 0.8 1.0 1.2
0.00 0.50 1.00 1.50 2.00 2.50
User Data Rate per Subchannel (Mbps)
Cumulative Distribution Function
2 × 2 OL MIMO 2 × 4 OL MIMO 4 × 2 OL MIMO 4 × 2 CL MIMO 10th Percentile Data Rate
12.4Summary and Conclusions
In this chapter, we provided some estimates of a WiMAX network system performance and its dependence on various parameters, such as frequency reuse, scheduling algorithm, subcarrier permutation, and MIMO. Based on these results, we can derive the following high-level conclu- sions on the behavior of a WiMAX network.
•Although a WiMAX network provides the highest per sector average throughput with a (1,1,3) frequency reuse, it does so at the cost of poor cell-edge performance. In order to achieve an acceptable level of user data rate at the cell edge, a (1,3,3) frequency reuse is required. If, however, sufficient spectrum is not available, a (1,1,3) frequency reuse with segmentation would also provide good cell-edge behavior.
Table 12.7 Total Throughput per Cell Site for Band AMC in a Ped B Multipath Channel with 30MHz of Spectrum (1,1,3) Frequency Reuse
Ped B Ped A
Profile DL (Mbps) UL (Mbps) DL (Mbps) UL (Mbps)
Basic configuration
(2 × 2 open loop) 131.47 21.13 139.88 22.48
Enhanced configuration 1
(2 × 4 open loop) 236.79 21.13 245.07 22.48
Enhanced configuration 2
(4 × 2 open loop) 200.34 32.20 209.33 33.64
Enhanced configuration 3
(4 × 2 closed loop) 306.99 49.34 315.99 50.78
Table 12.8 Fifth and Tenth Percentile Data Rates per Subchannel for Band AMC in a Ped B Multipath Channel with (1,1,3) Frequency Reuse
Scenario Ped B Ped A
5% (Mbps) 10% (Mbps) 5% (Mbps) 10% (Mbps) Basic configuration
(2 × 2 open loop) 0.025 0.085 0.025 0.085
Enhanced configuration 1
(2 × 4 open loop) 0.075 0.206 0.065 0.198
Enhanced configuration 2
(4 × 2 open loop) 0.035 0.095 0.035 0.090
Enhanced configuration 3
(4 × 2 closed loop) 0.437 0.499 0.371 0.482
•Scheduling algorithms and their ability to take advantage of multiuser diversity can lead to a significant improvement in the average throughput. We typically observe a 25 percent improvement in the cell throughput if multiuser diversity with a large number of users is used.
•Diversity, particularly at the receiver, provides significant gain in the average through- put—typically, 50 percent to 80 percent—but does not provide sufficient improvement in the cell-edge behavior to be able to use (1,1,3) frequency reuse without segmentation.
•Closed-loop MIMO with linear precoding and beamforming seems to provide a significant improvement in both the cell-edge experience and the average throughput. In going from an open-loop 2 × 2 MIMO configuration to a closed-loop 4 × 2 MIMO configuration, we observe an improvement of 125 percent to 135 percent in the per sector average throughput. The fifth and tenth percentile data rates are also significantly improved, indi- cating improvement in the cell-edge behavior. This would seem to indicate that with the closed-loop MIMO features of IEEE 802.16e-2005, a frequency reuse of (1,1,3) is usable.
•The overall spectral efficiency—defined as the average throughput per sector divided by the total frequency needed for the underlying frequency reuse—of a WiMAX network is quite high compared to the current generation of cellular networks. Even with a 2 × 2 open-loop MIMO configuration, a WiMAX network can achieve a spectral efficiency of 1.7bps/Hz with (1,1,3) reuse. The overall spectral efficiency in a pedestrian environment can increase to 3.9bps/Hz if closed-loop MIMO is implemented.
12.5 Appendix: Propagation Models
Median pathloss in a radio channel is generally estimated using analytical models based on either the fundamental physics behind radio propagation or statistical curve fitting of data col- lected via field measurements. For most of the practical deployment scenarios, particularly non- line-of-sight scenarios, statistical models based on empirical data are more useful. Although most of the statistical models for pathloss have been traditionally developed and tuned for a mobile environment, many of them can also be used for an NLOS fixed network with some modification of parameters. In the case of a line-of-sight-based fixed network, the free-space radio propagation model (see Chapter 3) can be used to predict the median pathloss. Since WiMAX as a technology has been developed to operate efficiently even in an NLOS environ- ment, we focus extensively on this usage model for the remainder of the appendix. We describe a few of the pathloss models that are relevant to NLOS WiMAX deployments.