Sunday, September 18, 2011

CAPACITY DIMENSIONING - 4G



Further to providing adequate radio coverage to customers, the next equally important objective is to ensure sufficient air-interface capacity (throughput) to offer a wide range of services. Given a business plan, the network capacity is driven by the potential subscriber number and the committed rates for data and VoIP services, as discussed in services dimensioning. The purpose of capacity dimensioning is to translate the capacity into number of sectors, which then have to be distributed for the estimated number of PoP. The missing parameter is average operating sector throughput, which finally determines how many subscribers can be served in a sector. Defining the average sector throughput is quite complex as it involves interpretation of a wide range of parameters such as
  • Accurate equipment specifications: PHY modes and SINR thresholds, advanced antenna system impact, interference mitigation mechanisms, scheduling intelligence, and terminal types.
  • Deployment considerations: Available spectrum, reuse factor and channel assignment method, coverage type(s), propagation environment, and site suitability.
  • Designer capabilities: Very good knowledge of the IEEE 802.16e standard, formal training or academic expertise in planning methodologies and their impact on practical networks, past deployment experience.
The sector throughput is usually provided as recommendation by the system vendor, however it can vary a lot depending on the deployment scenario. During dimensioning a more accurate estimation can be made through two approaches: 
(1) By extensive RF planning simulations and statistical processing that would result in a PHY mode regions map and give the average throughput. If clutter data are available this approach may provide a better view of the network performance. 
(2) By previous experience and interpretation of the above factors based on system performance observation. The former approach is time consuming, however it is much more accurate and is usually followed during actual network design. The latter is more appropriate during dimensioning.
Typically the equipment specifications include a table that matches the available PHY modes with their SINR requirement and the resulting Ethernet throughput (for different channel bandwidths) as in Table 1. The SINR values are given in the standard. The provided Ethernet rates correspond to a 60/40 DL/UL asymmetry and hence the TDD rate is the sum of DL and UL rates. The upper throughput bound can be achieved for an interference-free sector where terminals are located only in the 64QAM, 3/4 region, achieving around 10 Mbps TDD for the 5 MHz bandwidth. During practical deployments, the terminals will be scattered across the whole cell footprint, hence operating in various modes. Furthermore, interference due to frequency reuse may further downgrade the PHY mode for a particular terminal, especially if the number of channels is limited. A default assumption is to consider, as average sector throughput, the one corresponding to 16QAM, 1/2. Thereafter if the deployment conditions are favorable, as in the case of fixed-outdoor terminals or when enough spectrum is available for relaxed reuse, higher throughput should be expected. The throughput can be also enhanced by means of MIMO techniques and this should also be taken into account. It should be noted that the values in Table 2 refer to full usage of subcarrier (FUSC) permutation scheme, where all subchannels are allocated to users, hence the whole channel is exploited. In case segmentation is considered, i.e., partial usage of subcarrier (PUSC), the users in a sector utilize a specific segment (1–6 subchannels), and therefore the throughput in this case is reduced accordingly. Based on the standard, there will be regions in the DL and UL subframes for both FUSC and PUSC and in this case an average throughput condition should be expected. In most products, during the sector configuration, an RF designer can select or exclude segmentation according to deployment conditions.
Table 1: Ethernet Rates per PHY Mode (IEEE 802.16e System) 
  
5 MHz Bandwidth
10 MHz Bandwidth
PHY Mode (CTC)
SINR (dB)
DL (Mbps)
UL( Mbps)
DL (Mbps)
UL (Mbps)
QPSK 1/2
2.9
1.4
0.9
2.7
1.7
QPSK 3/4
6.3
2.2
1.3
4.3
2.7
16QAM 1/2
8.6
2.8
1.7
5.8
3.6
16QAM 3/4
12.7
4.3
2.6
8.6
5.4
64QAM 1/2
16.9
5.8
3.5
11.6
7.1
64QAM 3/4
18.0
6.5
3.9
12.9
8.2
The process of selecting frequency reuse and channel allocation in sectors is very important for both capacity and coverage. In mobile WiMAX this can be done in a flexible manner, although frequency planning cannot be avoided. This is due to the possibility of nonuniform network layout, in most cases, where frequency planning may improve performance. According to mobile WiMAX terminology the reuse is denoted as 1.x.y, where x denotes the cell sectors and the available channels. There are two main schemes under consideration: global reuse 1.x.1, where a single channel is used everywhere and cell/cluster reuse 1.x.y, where nx= 1, 2, 3. The most appropriate scheme is adopted, based on the systems’s special capabilities to reject or tolerate interference (i.e., via BF). An indicative performance of the most common  schemes for nomadic/mobile terminals (most sensitive to interference) is presented in Figure 1. It can be observed that for the 1.3.1 scheme the SINR drops well below the 3 dB threshold for the lowest modulation scheme, QPSK, hence a significant part of the cell footprint, especially among adjacent sectors, has no coverage. In the case of the 1.3.3 scheme, the interference appears closer to the cell edges and hence the coverage blanks spots are much smaller. It should be noted that in the 1.3.3 scheme the higher order PHY mode schemes extend to a larger region, hence indicating an improved sector throughput. Furthermore, when employing PUSC instead of FUSC, the 1.3.1 scheme behaves essentially as 1.3.3, while 1.3.3 as 1.3.9. It is typical for a sector to operate in 1.3.1 FUSC mode for terminals with good link quality and short link distance and in PUSC mode, which is equivalent to 1.3.3 for terminals that would otherwise achieve low SINR (due to low signal strength or interference).

Figure 1: SINR map for (FUSC) and (PUSC) schemes.
Knowing the average sector throughput as described in previous paragraphs, capacity dimensioning can be completed as follows: Initially an analysis on the customers that can be accommodated in a sector is performed. This is done by analyzing the service plan and calculating the average data and VoIP CIR per service and customer. Then a graph that shows the throughput versus the subscribers is obtained, as in Figure 2. The average sector throughput is projected in the sector total graph indicating the corresponding subscriber number. Considering a 10 Mbps throughput, each sector can accommodate around 70 customers. Moreover a graph such as in Figure 2 is quite useful also during network operation as when the throughput takes a higher value, due to special circumstances, additional subscribers can be served. The second step is to divide the number of customers per area with the customers/sector to calculate the required number of sectors.

Figure 2: Estimated sector and services throughput versus subscribers.

No comments:

Post a Comment

Related Posts with Thumbnails