Praise for Fundamentals of WiMAX

12.3 System-Level Simulation Results 413manually oriented in order to get a strong signal. Most WiMAX desktop devices are expected tobe equipped with six to eight such antennas, each with a gain of 3dBi to 6dBi.Figure 12.3 and Figure 12.4 show the average throughputs per sector for the basic configurationin Ped B and Ped A environments, respectively. The average throughput per sector isslightly better in the case of a Ped A channel than in a Ped B channel because the Ped A channelprovides better multiuser diversity due to larger variations in channel amplitude, which isexploited by the proportional fairness scheduler.The overall per sector throughput in the case of (1,1,3) reuse is better when a directionalantenna is used at the MS, since the amount of cochannel interference is reduced by the directionalnature of the channel. However, in the case of (1,3,3) reuse, the additional directionality ofthe antenna at the MS in an interference-limited environment does not provide any significantbenefit, since (1,3,3) frequency reuse provides a sufficient geographical separation of cochannelBSs. It should be noted that in the case of noise-limited design—a design with larger cell radii—the gain of the directional antenna at the MS would provide an improvement in the sectorthroughput even with (1,3,3) frequency reuse.Figure 12.5 and Figure 12.6 show the probability distributions of per subchannel user DLdata rate for the Ped B and Ped A environments, respectively. One can conclude that in the caseof (1,3,3) reuse, the fifth and tenth percentile data rates are much higher than the case of (1,1,3)reuse. This happens because in the case of (1,1,3) reuse, a large percentage of MSs that arepresent toward the cell edge experience a low SINR, due to cochannel interference and thus alow data rate. Based on the per user data rate distribution it should be noted that although (1,1,3)reuse is more spectrally efficient, it is achieved at the price of poor performance at the cell edge.In order to achieve an acceptable cell-edge performance, (1,3,3) reuse or (1,1,3) reuse with segmentationis required. When segmentation is used, all the subchannel are divided into threegroups, and each of the three sectors is allocated one group of subchannels. Segmentation thusachieves an effective (1,3,3) reuse.The fifth and tenth percentile data rates can also be improved in the case of (1,1,3) reuse byusing directional antennas at the MS, as shown in Figure 12.6. This controlled trade-off betweennetwork reliability and spectral efficiency allows a system designer to choose the appropriatenetwork parameters, such as cell radius, frequency reuse, and antenna pattern, that will meet thedesign goal. From here on, we limit our discussion to the handheld-device scenario with (1,1,3)frequency reuse.Table 12.4 and Table 12.5 summarize the throughput per BS and the fifth and tenth percentiledata rates for the various scenarios. The throughput of all the sectors is combined to get thethroughput of the BS. Since a total of 30MHz of spectrum is assumed, as per Table 12.2, in thecase of (1,1,3) frequency reuse, we assume that each sector is allocated three 10MHz TDD channels.Although the average throughput channel is less in the case of (1,1,3) frequency reuse thanfor (1,3,3) reuse, the overall capacity is higher with (1,1,3) reuse, since each sector is allocatedthree channels as opposed to one channel in the case of (1,3,3) reuse. On the other hand, networkreliability is significantly improved by going from (1,1,3) reuse to (1,3,3) reuse.