3

Planning of WiMAX and LTE Networks

Mehrnoush Masihpour and Johnson I Agbinya

University of Technology, Sydney Australia

PART II
Planning and Dimensioning of WiMAX Networks

3.1 WiMAX Network Planning Parameters

The key air interface technology used in WiMAX is OFDM. OFDM is an efcient form of multi-carrier modulation schemes, which is used in most of the modern communication technologies, such as 3G-LTE, DSL, 4G Cellular Systems and WiMAX. To understand OFDM we introduce a few pertinent terminologies that are specic to it. OFDM system bandwidths are normally purchased by telecommunication operators through spectrum bids. The main system bandwidths range from 1.25 MHz, 3.5 MHz, 5 MHz, 8.75 MHz (WiBRO), 10 MHz and 20 MHz. Except for a small region of the spectrum (guard band), the rest is used for carrying data signals and for synchronization of the receiver to the transmitter. The FFT therefore occupies the spectrum in the frequency domain. The bandwidth is given by the expression B = Nc(max) . f (3.1) Where B is the OFDM signal bandwidth and f is the subcarrier spacing in Hz. This bandwidth does not include the guard band area. OFDM analog
67

68 Planning of WiMAX and LTE Networks to digital converters sample the incoming signal at a rate which permits for adding guard bands (cyclic prex). OFDM is sampled using a so-called sampling factor. The sampling factor is the ratio of the sampling frequency to the OFDM bandwidth. Usually OFDM is sampled at a frequency a bit larger than the critical Bandwidth to make provision of guard bands. The sampling factor is: Fs (3.2) OF DM bandwidth Where Fs is the sampling frequency. Some of the common sampling factors are 8/7, 28/25, 78/75, The available spectrum is sampled using the equation (3.3): n= FS = Floor n.BW 8000 .8000 (3.3)

Where Floor implies rounding down the result of the equation and n is the sampling factor and depends on the available spectrum (bandwidth). For WiMAX the channel spacing is f = FS /NFFT , where NFFT is the size of the FFT used to implement WiMAX. For example in xed WiMAX NFFT = 256. The symbol time is given by the relationship TS = 1 (1 + G) and G = m ; m = {2, 3, 4, 5} f 2 (3.4)

3.2 OFDM Symbols

To understand how OFDM operates, in [8] some of the important terms are dened [8] and their relationships. Channel delay spread ( ): the time difference between the arrival of the rst multipath component and the last one. Useful Symbol Time (Tb ): The information bearing signal samples occupy the useful symbol time given by the relationship Tb = 1f . This time does not include the guard band time. Guard Period Interval/Ratio: To provide for multipath effect, a guard time proportional to the anticipated multipath is added to the useful symbol time. The guard time is related to the useful

3.2 OFDM Symbols

69

symbol time by the relation: Tg = G Tb , where G = 21 ; m = m {2, 3, 4, 5}. The cyclic prex samples occupy this time period. Overall OFDM Symbol time (Ts ): This is the time over which an OFDM symbol is valid. The symbol time is given by two times, the guard band time and the useful symbol time over which the actual information bearing signal samples are present. Therefore the total time over which the OFDM FFT is performed is given by the sum of the guard band time and the useful symbol time: TS = Tb + Tg = (1 + G)Tb or TS = (1 + G) f (3.5)

Coherence bandwidth (Bc ): the bandwidth in which the channel is assumed to experience a at fading propagation and it is approximately equal to 2/ If Ts , number of symbols per second is high, which means high data rate transmission. In this case, the system is very fragile to ISI and it is desirable to minimize the ISI. Therefore, to address this problem, multicarrier modulation schemes reduce the Ts for the individual sub-bit streams to have [8]. Ts The raw capacity of the channel per symbol can therefore be estimated and is: k.Nsubcarriers (3.6) Craw = TS Where k is the modulation order and different modulation schemes are used to cover different areas of a cell as explained in Figure 3.1 of Chapter 3. For a 3.5 MHz bandwidth, the symbol time neglecting the guard band is 64s and if G = 1/32, the total symbol time is 66s. With 192 subcarriers and using 64QAM modulation with 6 bits per symbol, the capacity is: 6x192 = 17.45 Mbps 66s This is a theoretical result which assumes that the transmission channel does not introduce errors. In practice transmission channel errors are introduced and forward error correction (FEC) is required to safeguard against Craw =

70 Planning of WiMAX and LTE Networks

Tg
Cyclic Prefix

Tb
Data Payload

TS
Fig. 3.1 OFDM symbol time. Table 3.1 Mandatory FEC coding rates per modulation. Modulation BPSK QPSK QPSK 16-QAM 16-QAM 64-QAM 64-QAM Uncoded block size (bytes) 12 24 36 48 72 96 108 Coded block size (bytes) 24 48 48 96 96 144 144 Overall coding rate 1/2 1/2 3/4 1/2 3/4 2/3 3/4 Rs code (12, 12, 0) (32, 24, 4) (40, 36, 2) (64, 48, 8) (80, 82, 4) (108, 96, 6) (120, 108, 6) CC code rate 1/2 2/3 5/6 2/3 5/6 3/4 5/6

errors. FEC is implemented by using redundant bits in each symbol. This is an increase in the bits used to represent the symbol. The coding rate for the FEC is the ratio of the number of information bearing bits to the total number of bits including the redundant bits. The mandatory FEC rates for different modulation schemes are given in Table 3.1. The effect of FEC is to reduce the raw capacity by the coding rate. Thus we have CR k Nsubcarriers (3.7) C = Craw x CR = TS Thus in the previous example the useful capacity when CR = 3/4 becomes C = (17.45 3)/4 = 13.1Mbps. As a measure of how effective the 3.5 MHz bandwidth has been used, the spectral efciency is given: S = C 13.1 Mbps = = 3.74 b/ sec /H z B 3.5 MH z (3.8)

The spectral efciency improves with higher coding rates. In other words, the less the number of redundant bits used for forward error correction, the better the spectral efciency. The theoretic WiMAX capacity was estimated based on this understanding and with a 20 MHz bandwidth, symbol time of 11.3s, 192 used subcarriers

3.3 WiMAX Frame

71

and (k) 6 bits per symbol at CR = 3/4 as: C= 0.75 192 6 CR k Nsubcarriers = 76.46 Mbps = TS 11.3s (3.9)

This theoretical WiMAX capacity estimate overlooks the overheads from the MAC layer and PHY. It also fails to highlight the fact that the range of a base station with such a capacity will be a lot shorter than often advertised for WiMAX by many authors. The range 70 km or more is normally for BPSK modulation and the large capacity is for 64QAM modulation and not distinguishing the two often leads to wrong conclusions about WiMAX capacity versus range. This confusion is created not by the standards body but by various writers on WiMAX.

3.3 WiMAX Frame

WiMAX symbols are normally sent in groups of symbols called a frame. Rather than sending the symbols as a stream, they are formatted into a time domain multiple access (TDM) frame. In practice the symbol time varies with the width of the channel. Therefore whole numbers of symbols do not necessarily t into a frame snugly. There are often small gaps at the ends of frames that occur and not used. This overhead is often less than a symbol period in a frame and wastes capacity. The impact of this unused gap is more prominent in small frames. As specied in the standard frames may be of lengths TF = [2.5, 4, 5, 8, 10, 12.5, 20 ms]. The number of symbols per frame N is: N = FLOOR TF TS (3.10)

For example in a 3.5 MHz channel with a cyclic prex (CP) of 1/8, the symbol time is 72s and the number of symbols in 8 ms frame is N = FLOOR 8 103 72 106 = FLOOR 1000 9 = 111 symbols

The unused gap at the end of the symbol is TF (111 72s) = 8 ms 7.992 ms = 8s, in this case a small reduction in the capacity (0.1% reduction).

72 Planning of WiMAX and LTE Networks

3.4 Multicarrier Modulation schemes

In modern wideband high data rate transmission systems such as WiMAX, the effect of the ISI is very crucial, because in this technologies the delay spread Ts . In WiMAX this can easily happen when the range of the cell is large enough to cause signicant multipath delay spread. The result is more ISI and consequently more errors to occur. Therefore modern communication technologies tend to use multicarrier modulation schemes in the physical layer to reduce the ISI. In the modulation schemes, a high data rate bit stream is divided into L lower rate sub-streams and transmitted on L parallel sub-channels or sub-carriers, which are orthogonal to each other in an ideal channel. The symbol time for each sub-stream is Ts /L , thus the ISI will be decreased. Although the data rate for each sub-channel is equal to RTotal /L, the total required data rate will be retained as all sub-channels are parallel to each other. Similarly the bandwidth of each sub-channel is derived by Btotal /L. To overcome the ISI in each sub-channel with a at fading propagation, the bandwidth of each must be far less than the coherence bandwidth (Btotal / L Bc ). It can be derived from this relationship that the more the number of sub-carriers, the less bandwidth and data rate for each sub-stream but the more the sub-streams symbol time and resulting in less ISI. Although, there are some limitations associated with the multicarrier modulation such as high cost of low pass lters to achieve orthogonality and the requiring of a number of independent frequency channels for each sub-streams, OFDM technique copes with those problems and efciently uses the available spectrum. 3.4.1 Guarding Against Inter-Symbol Interference OFDM is well known for creating Inter Symbol Interference (ISI) free channels [8]. The interference between a symbol and the subsequent symbols is referred to as ISI, which is an undesirable effect of the environment, caused by either multipath propagation or Doppler effect. When a signal reaches the destination from different paths due to reection, refraction and scattering, the propagation process is called multipath propagation. Dividing a high data rate stream of bits into several lower data rate streams and transmitting each bit stream via parallel individual subcarriers is the technique that multicarrier modulation scheme uses for data transmission [8].

3.4 Multicarrier Modulation schemes

Required orthogonality interval Orthogonality interval

73

T1 T2
0 Time

Symbol 1

Symbol 2

T3
Delay ISI Symbol 1 ISI Symbol 2 ISI 0 Required Orthogonality interval Time

Orthogonality interval

T1 T2
0 Time

Symbol 1
Guard interval

Symbol 2

T3
Delay 0 ISI Symbol 1 ISI Symbol 2 ISI Time

Fig. 3.2 Explanation of OFDM guard intervals and prex (Adapted form [9]).

Radio signals often propagate from the transmitter to the receiver through many paths due to obstructions from objects in the terrain (the space separating the transmitter and receiver). Hence several copies of the same transmitted signal arrive at the receiver time shifted from each other. The time shifts are proportional to the paths taken by the signal and the effects of the channel (see Figure 3.3) on the signal. Assume that the direct path signal arrives at the receiver rst. Thus delays encountered by signals arriving later can be estimated relative to the time of arrival of the direct path signal. The effect of adding the different copies of the received signal at the receiver causes signal power degradation (known as fading) due to intersymbol interference as shown in Figure 3.3. In this Figure three components of the same signal are received with the channel delaying them by T1 , T2 and T3 seconds respectively.

74 Planning of WiMAX and LTE Networks

Required orthogonality interval Orthogonality interval

T1 T2
0 Time

Symbol 1

Symbol 2

T3
ISI Delay 0 Symbol 1 ISI Symbol 2 ISI Time

Fig. 3.3 Intersymbol interference at receiver [9].

Two separate regions can be identied in the received signals. Firstly, one due to time shifted version of the same symbol, called self-symbol-interference (SSI) and secondly interference from neighbouring symbols called intersymbol interference (ISI). OFDM resists SSI because it uses orthogonal frequencies in the transmitter leading to zero correlation for two different frequencies. Hence SSI can be constructive in OFDM systems as the signal in such regions produce higher signal to noise ratio. This advantage is more prominent with the use cyclic prexes, which ensures SSI is removed. ISI is provided for in OFDM systems through the use of the guard period (orthogonality interval, Figure 3.3). The guard band is proportional to the maximum delay expected for all the paths the signal could take to arrive at the receiver. The guard period is implemented by adding the last part of an OFDM symbol to its front as a guard against the delay expected on arrival. Thus the desired signal could arrive at the receiver without error and at a good time distance away from copies of the same signal that have been received. What should the length of the guard period be? The length of the guard period is variable and can only be assumed and it is proportional to the largest delay path the signal could take from transmitter to receiver. A good estimate of this value requires a simulation of the channel (path) the signal should take from transmitter to receiver. Thus information about the nature of the terrain separating them is required. If the reection paths are long, the guard interval

3.4 Multicarrier Modulation schemes

Table 3.2 Example of system parameters. Parameter Bandwidth Sampling factor Sampling frequency Size of FFT Subcarrier spacing Useful symbol time The guard intervals Cyclic prex time delay path (given by the speed of light) Overall symbol time (G=1/4) Number of Symbols Length of subframe Length of Frame Number of symbols per frame User subcarriers Pilot subcarriers Used data subcarriers Modulation (QPSK), bits in subframe Raw bit rate (no coding) Variable and Calculations B n FS = n.B NFFT f = FS /NFFT Tb = 1/ f G
1 Tg = GxTb = 32x 4

75

Value 7 MHz 8/7 8 MHz 256 31.25 kHz 32s 1/4, or 1/32 8s or 1.0s 2.4 km or 0.3 km 40s 25 1000s 10 ms 250 200 8 192 9600 bits 9.6 Mbps

ddelay = cxTg = 3x108 x8s ddelay = cxTg = 3x108 x1s TS = Tb + Tg = 32s + 8s NSymbols Tsym.frame = NSymbols xTS = 25x40s TF N = FLOOR 10 ms 40s Nuser NPilot NUsed = NUser NPilot Nbit = NUsed x25x2
9600 Rb(raw) = 1000s

should also be long and vice versa. There is a penalty in using large guard times as it causes a reduction in throughput due to wasted bandwidth. Long range base stations in suburban areas should use larger guard interval. For example suppose the system parameters are as in Table 3.2. 3.4.2 Overcoming Doppler Shift A moving radio transmitter relative to a receiver or vice versa causes the transmission frequency to change. This change in frequency and its effects on the received signal power is termed Doppler effect. Reducing the distance between the transmitter and receiver causes the carrier frequency to increase and increasing the distance between them causes the frequency to decrease. In an OFDM system when the transmitter and receiver move closer the carrier frequency increases and when they move apart it reduces. The change in carrier frequency is proportional to the relative velocity between the transmitter and receiver. This change is given by the expression: f = v fC c (3.11)

76 Planning of WiMAX and LTE Networks Where, fC is the carrier frequency and c is the speed of light. Consider a fast moving car at 120 km/hr carrying a terminal point transmitting at 6 GHZ. The Doppler shift is as high as 670 Hz. This change is substantial for a carrier frequency of 31.25 kHz (a 2.144% change in carrier position). Thus the demodulator at the receiver must be able to track this change and make corrections for it.

3.5 General Considerations for WiMAX Network Planning

Before the establishment of a communication network, a careful planning and dimensioning of the network is necessary. Through this process, network designers gather the analytical data and using practical design methodologies to obtain an optimum network design to increase the network performance, while reducing the time and expenses required for network implementation [1]. Network dimensioning and planning is the most signicant part of the network design, which should be considered from three different perspectives of dimensioning for service, coverage and capacity. To achieve a proper planning, required information and Key Performance Indicators (KPI) need to be well dened. Some of the most important information that has to be identied is as follows [1]: Geographical area, which is due to service delivery, needs to be identied in terms of the size in km2 also the prole that denes the area as either urban, suburban or rural environment. Furthermore the coverage area should be distinguished in terms of the network being xed, nomadic, portable, and mobile or any combination of them. Subscribers prole such as residential, small businesses and corporation customers needs to be identied. How subscribers are distributed through the network, which can be based on different parameters such as the number of the users per a service area, per a particular prole or per a period of time. Service prole is also very important, particularly for the service dimensioning. VOIP, broadband Internet access and IPTV are some of the services that WiMAX can offer. Available spectrum, in which WiMAX network is allowed to communicate within.

3.6 WiMAX Cells: Cellular Technology

77

Cartographic data to illustrate the service area through a digital map. KPIs as a measure of performance Customer requirements such as bandwidth, number of base stations and type of site. Using the information outlined above, a WiMAX network planning and dimensioning for service, coverage and capacity can be done.

3.6 WiMAX Cells: Cellular Technology

Cellular technology is used by most of the modern communication systems such as GSM and UMTS. WiMAX also follows a cellular pattern to address the need for increasing the capacity. To understand how WiMAX technology works, it is necessary to become familiar with the main concepts of the cellular networks. In this respect, cellular technology is discussed in some detail in this section. The growing number of subscribers requiring access to the communication networks in one hand and limited available radio frequency spectrum on the other hand, consequently the need for more capacity, has led the ICT (Information and Telecommunication Technology) scholars to develop the cellular system to allow the allocated radio frequency to be reused. A cellular network refers to a radio network consisting of a number of cells (dened by the geographical coverage of a base station), each one serving a specic geographical area by means of a base station through the air interface. In other word, a geographical area, which requires the coverage for the data transmission, is divided into a number of cells and each cell allocated a set of radio frequencies. Although, to minimize the interference between the frequencies, none of the adjacent cells can use the same set of frequencies, the cells far enough from each other can be allocated the same frequency band [6]. This is referred to as frequency reuse, which is discussed in some detail in the Section 1.3.2. 3.6.1 Cell Shape and Size A hexagon is usually assumed to be the shape of a cell theoretically, however, in practice it is amorphous due to the environmental impact on the signal propagation and also the technology in use such as the type of antenna applied

78 Planning of WiMAX and LTE Networks

(a)

(b)

Fig. 3.4 (a) Circular cells with blind spots; (b) Overlapping circular cells.

Fig. 3.5 Cell shape.

at the base station and its location. Figure 3.4(a) illustrates an ideal cell shape. Other possible shape of a cell is either a circle or square [6]. If the geographical area is divided into a number of circles, the gap between them is supposed to be a blind spot therefore it has an improper shape for a cell. In Figure 3.4(a) there are blind spots between the circles. Including the purple circle in the middle does not completely remove the blind spots. To remove the blind spots, the cell footprints must overlap as in Figure 3.4(b) which results to wasting of resources and coverage. A square is also inappropriate because a square cell has unequal distance with its neighbors [6]. If we assume a square width of d it has four neighbors (Figure 3.6) at distance d and four at distance 2d therefore it requires complicated computations for some required algorithms also antennas using in a cell should be none equidistant.

3.6 WiMAX Cells: Cellular Technology

d d d d d d R d d

79

d d

Fig. 3.6 Comparison between the cell shape and computations/square and hexagon.

A hexagonal cell, having equal distance with all adjacent cells provides the best coverage model with the least number of required base stations to serve a particular area. Assuming R as the cell radius, the distance between two adjacent cells is d = 3R and the coverage area of a cell is approximately 2.598R 2 . However, we also the dene the size of a cell in terms of its capacity and referred to as the number of users which can use a cell simultaneously. According to the cell size, there are three different cells. Macro cell, which is large and has radius between 100 m to 10 Km or more, is suitable for the coverage in rural area with the low number if users. Micro cells are the medium size cells and have radius of about several 100 m, this type of cells are proper for crowded urban areas such as shopping malls. The smallest cells are referred to as Pico cells, which have size about a few dozen meters, can be usually implemented in networks in areas such as ofces or in lifts. As the size of the cell decreases, the required transmission power also decreases and users need less transmission power. The height of the base station antenna is lower for the smaller cells. 3.6.2 Cluster and Frequency Reuse To increase the capacity of the network, while using a limited number of radio channels, cellular systems use the frequency reuse technology. It means that the same number of frequencies is used as many times as required in a network by imposing a careful frequency plan. A cellular network consists of at least one cluster. A number of cells, each one using different group of frequencies from the other cells, form a cluster. The coverage range of a cluster is known as a footprint. Different cells

80 Planning of WiMAX and LTE Networks

2 3 1 4 5 6 7

Fig. 3.7 Cluster with K = 1/7.

associated with a cluster must be allocated a specic and unique subset of all the available duplex channels to prevent interference between the same frequencies. Assuming S as all the available pair of channels and N as the number of channels used by each cell, the number of cells in a cluster is deduced by K = S/N. Inverse of this number of the cells is referred to as frequency reuse factor (1/K). The possible values for Kare 1, 3, 4, 7, 9, 12, 13, 16, 19, 21, etc [6]. Figure 3.7 shows a cluster with K = 1/7. A cluster can be repeated in the network as many times as needed. If we assume the cluster to be repeated M times, the total capacity of the network will be C = M S, which is the total number of the radio channels available to be assigned to the subscribers at the same time. To prevent the interference between the co-channel cells, which are the cells using the same group of radio frequencies, they need to be far enough from each other. The minimum required distance between the co-channel cells is calculated by: Assuming the base stations in the co-channel cells are located in C(u1 , v1 ) and F(u2 , v2 ) and distance between them is D derived by: D = {(u2 u1 )2 (cos 30 )2 + [(v2 v1 ) + (u2 u1 ) sin 30 ]2 }1/2 D = {(u2 u1 )2 + (v2 v1 )2 + (v2 v1 )(u2 u1 )}1/2 (3.12)

We restrict (u1 ,v1 ) to integer values (i,j) and set (u1 , v1 ) = (0, 0) to give the expression D 2 = i 2 + ij + j 2 (3.13)

3.6 WiMAX Cells: Cellular Technology

81

A A D

A A A

A A

Fig. 3.8 Distance between co-channel cells.

In order therefore to connect without gaps between adjacent cells, the number of cells per cluster (cluster size), D, must satisfy: D 2 = i 2 + ij + j 2 (3.14)

Where i, j are non-negative integers. To nd the nearest co-channel neighbours of a particular cell: (1) Movie i cells along any chain of hexagons (Figure 2.6) and (2) Turn 60 degrees counter-clockwise and move j cells.

3.6.3 Channel Assignment in Cellular Systems Each cell can use a specic number of channels for allocation to its users. Channel assignment, which means designation of the available channels to the base stations or to the users calls, requires a precise planning to minimize

82 Planning of WiMAX and LTE Networks the possible interference to enhance the capacity of the network. To achieve that there are three main strategies, static or xed, dynamic and a hybrid form of both. In the static channel assignment, which suites the network with the uniform trafc and high total trafc load in all cells, all the available channels are allocated to the cells and each base station has its own number of channels [7]. In this pattern, if all the channels are in use, new call request will be blocked. However, in dynamic channel assignment, the channel will be allocated to a call at the time that the call is made. In other word, no base station has its own specic and constant number of channels. A combination of two is used in the network with multiform trafc load but slightly the same trafc ratio, which means the trafc load have not many changes over time. To implement a hybrid channel assignment, all the available channels are divided into two sets of static and dynamic channels. In the case that the cells have different trafc load, three main strategies can be done to prevent the calls to be blocked. Borrowing the available channels from adjacent cells is a solution for serving a user, while all channels in a cell are occupied by other users. In this case the borrowing channels are released to the original cell when the call is terminated. Another way to overcome this problem is to allocate the channels to the cells based on their needs. In this case, all the available channels are maintained in a pool and are assigned to the cells when needed. Channels can also be allocated by a central decision element, which decides allocation of the channels; however it can slow down the network. As the number of users grows the capacity available to a user decreases. Alongside, frequency borrowing from adjacent cells, as a method of increasing the networks capacity, cell splitting (Figure 3.9(b)) also can be done to achieve this goal. Cell splitting refers to minimizing the cell size. In this scheme, a cell is divided into smaller cells, therefore the number of cell increases consequently the number of required base stations is also boosted and he achieved capacity is much higher. This scheme operates properly in crowded areas such as big cities, particularly in busy spots such as shopping malls. Another method of increasing the capacity is cell sectoring. Using directional antennas at the base station within a cell, instead of using Omnidirectional antenna, is referred to as cell sectoring (Figure 3.9(a)). Three 120 or six 180 directional antennas are usually deployed in a base station to sector a cell. This leads to less co-channel interference but more intra cell hand over.

3.6 WiMAX Cells: Cellular Technology

83

Fig. 3.9 Cells sectoring (a) and splitting (b).

These schemes are used in WiMAX networks with some unique variations which are discussed in latter sections of the chapter.

3.6.4 Handoff (Handover) One of the most important operations in the cellular system is the concept of handover or handoff. When a mobile user changes the base station in use, which means moving from one cell to another, handover occurs. As it is described in the previous sections, different cells use different frequency channels, therefore when a user changes the cell and consequently is assigned channel the call would be dropped. Dropping a call is undesirable for a communication; therefore handoff solves the problem and maintains the connectivity while changing the cell. However, the handoff is unrecognizable to the mobile phone user. The reason for handoff is achieving a stronger signal for a desirable communication quality. When the receiving signal is not strong enough, the mobile station requests a handoff and the call will be switched to the new base station with higher signal level. There are two groups of handoff processes called horizontal and vertical handoff. Vertical handoff occurs when a mobile devices changes its point of attachment from a network of one type to another (eg. From a GPRS network to a UMTS or from a WLAN to UMTS etc). Handoff can occur between networks owned by different operators if commercial agreements exist between them to permit it.

84 Planning of WiMAX and LTE Networks Horizontal handoff occurs within the same network type (eg. within a UMTS network). There are several forms of handoff depending on the network conditions including a) Handoff between neighboring base stations (BTS) or cells b) Handoff between two RNCs c) Handoff between SGSNs This is important to perform the handover within less time as possible. To do that, three important factors should be identied, which are as follows: Minimum usable signal level for required quality. A threshold signal level An optimum signal level for a handoff to be performed. Having the above factors handover margin can be dened as: = P r(handover) P r(min
imumusable)

(3.15)

The handoff margin should be small enough to obtain optimum the handoff. Because unnecessary handovers occur if the is too large, whereas too small handover margin leads to the call disconnection, as there is no enough time for switching between the cells and weak signal can drop the call. To decide on handoff, different performance metrics can be used. Some of the metrics are dened below [6]: Cell blocking probability: the probability of blocking a new call, when the trafc load is very high and the base station cannot serve the call. Cell dropping probability: the probability of disconnecting a call in order to perform the handover. Cell completion probability: the probability of being an active call terminated successfully. Probability of a successful handover: the probability of performing a handover successfully Handover blocking probability: the probability of an un successful handover Handover probability: the probability of a handover to occur before the call is terminated

3.6 WiMAX Cells: Cellular Technology

85

Rate of handover: the number of handovers occurring per a period of time There are two types of handovers, soft handover and hard handover. Hard handoffs are operated when the connection is broken with the current base station before connecting to the target base station. However, if the connection to the new base station is performed before breaking with the current base station, soft handoffs occurs [6]. 3.6.5 Trafc Engineering Trafc engineering refers to optimizing the network performance by determination of the trafc patterns, predicting the growth rate and data analysis. In this section, trafc engineering is discussed based on the network capacity such as predicting the required capacity and data analysis on capacity issues. As explained earlier, the capacity of a cell is determined by the number of available channels within a cell which allows the same number of simultaneous users per cell. However, it is unrealistic to have the capacity to serve all the subscribers at the same time. In practice, an expected number of simultaneous users will be identied for the frequency planning and channel allocation purposes, because not all subscribers request for calls at the same time [6]. Assuming the call capacity to be N with S subscribers, the network is described as a blocking system if S > N and a non-blocking system if S < N . In practice systems are usually blocking systems. In case of a call being blocked two scenarios may happen based on the network conguration [6]. The call might be waiting in a queue to be served and the call might be dropped and rejected. In the second scenario, a cellular system assumes that the user hangs up and replicates the call after a certain amount of time (LLC model); however, another assumption is that user requests to call repeatedly (LCH model) [6]. For a blocking system to perform efciently, there are some factors to be considered such as the call blocking probability and average delay, if the blocked calls are waiting to be served by the system. Trafc is determined as trafc density (A), which is stated as a dimensionless unit in Erlang. The trafc density is the result of an average holding time per successful calls (h) multiplied by the average rate of call request per unit of time (), or it can be

86 Planning of WiMAX and LTE Networks dened as the average number of call requests during an average holding time. A=h (3.16)

The network should be deployed in the way that can support the highest possible trafc. The highest possible trafc occurs during the busy hour of the day. To estimate the highest possible load, according to the International Telecommunication Unit-Telecommunication (ITU-T), the average trafc load of transmitted during the busy hour of 30 busiest days of the year should be identied. In trafc engineering, it is important to consider the number of users as nite or innite. Assuming a nite number of users is acceptable if the number of users is at least, 5 to 10 times of the capacity of the network. In this case the arrival rate is constant. However, based on the assumption of innite number of users, the arrival rate will be dependent on the number of users, which are using the channels at the time. The arrival rate at time t is expressed as: ArrivalRate = (S K)/S Where: S is the total number of users Each user average rate of call is /S K is the simultaneous number of user at time t However, based on the assumption that the number of users is innite and the network follows a LLC pattern, the important parameter, the Grade of Service (GoS) Probability or call blocking probability (P ) should be considered. GoS determines the probability of a call to be blocked during the busy hour. The grade of service values between 0.01 and 0.001 are good values. Assuming trafc density of A and N number of channels, the grade of service is derived by the expression: P =
AN N! N AX N! x=0

(3.17)

(3.18)

From the equation, given the capacity of the system, the number of trafc can be determined for a given value of grade of service. Also given the amount of trafc, the capacity is identied for a certain grade of service.

3.7 Dimensioning for Service

87

3.6.6 Benets and Drawbacks of Cellular Technology To conclude the overview of cellular system, some advantages and disadvantages of this technology are summarized in this section: Advantages: It is easily possible to increase the capacity and number of users through creating new cells, sectoring, cell splitting and or by simply scaling the network up progressively Lower transmission power Increasing the coverage area Robustness Disadvantages: The major disadvantages include The need to engineer necessary handover within (and between) the networks Handoff is necessary Inter cell interference occurs due to co-channel interference and adjacent channel interference Fixed base stations are required. Newer versions of networks on platforms or vehicular networks seek to solve this limitation.

3.7 Dimensioning for Service

In the process of service dimensioning, different types of services offered to different customer proles is determined and also how it affects the air interface is analyzed. Typically, there are three important services offered by WiMAX which are VoIP, broadband data and guaranteed bandwidth [1]. According to the type of the service, a portion of the total available bandwidth in each sector is allocated to that particular type of service to deliver the QoS required by WiMAX. As described in Section 1.1.4, there are three different categories of applications required divergent QoS known as Unsolicited Grant Service (UGS), Polling Service (PS), Best Effort service (BE) [2]. VoIP is categorized as a PS while broadband data which corresponds to the Internet access is an application and requires BE service and guaranteed bandwidth is equivalent of UGS.

88 Planning of WiMAX and LTE Networks Usually a xed portion of the bandwidth identied as 20% [1] of the total bandwidth is assigned to the UGS during the primary conguration of the network. However, the remaining channel bandwidth is allocated dynamically to the BE and PS, each one a specic portion according to their needs. The impact of three different services on the WiMAX network is analyzed in the following sections. 3.7.1 Voice Over IP Service VoIP refers to the vocal communication delivered in a packet-switched network. In WiMAX technology, the voice stream is broken up into a number of data packets and transmitted via air interface. One signicant task in WiMAX planning is to estimate the capacity needed for delivering the VoIP service to subscribers, since any defeat to a desirable resource allocation results in degeneracy of the grade of service and subsequently degradation of several calls in the sector. To achieve that, some important characteristics of the voice service needs to be identied within a sector, such as total lines, active lines, grade of service, trafc activity and required data rate per call [1]. Mostly the required capacity per sector is needed to be calculated for a given number of simultaneous calls during the busy hour. Using Erlang B equation or table the number of active call will be identied for a given GoS: Pb = Where A: Sector trafc activity N: number of active calls Pb : blocking probability or grade of service (GoS) 3.7.1.1 Broadband data service This service offers access to data at high rates. It is evaluated by the peak information ratio (PIR) in Mbps, which is dened as the maximum data rate that a network can offer to the subscribers if the network is not congested [1]. However, since the broadband data service is a BE service, data rate will be lower when the number of simultaneous subscribers increases during
AN N! N AX x=0 X!

(3.19)

3.8 Dimensioning for Coverage

89

the busy hour known as data over-subscription rate (O). The ratio of peak information ratio Ppir to data over-subscription ratio is a performance indicator that expresses the committed information ratio, Pcir , which is the required data rate during the busy hour. Pcir = Ptotal Ppir O = (Pcir,i Ns,i ) (3.20) (3.21)

Ns,i : corresponding number of subscribers Ptotal : Summation of the per prole rates Ppir : proles CIR The rate for FDD is derived from the Max (DL, UL) and for TDD as the sum of DL and UL rates [1].

3.8 Dimensioning for Coverage

Through the coverage dimensioning process, the number of base stations required for coverage of a given area in km2 is identied in such a way that meets the satisfaction of KPIs [1]. The footprint of each cell needs to be calculated for the estimation of the number of base stations, Number of required base stations = Service Area(km2 ) Cell Footprint(km2 ) (3.22)

Cell footprint varies from different scenarios and equipment conguration. It refers to the maximum range that a base station supports to radiate the electromagnetic waves to achieve a performance threshold based on the signal power at the receiver. The received signal strength can be calculated using the following formula [1]: S(dBm) = P + G + Gsp Ld Lp M Where: S: Signal strength at receiver P : power of transmission G: antenna gain Gsp : signal processing gain (3.23)

90 Planning of WiMAX and LTE Networks Ld : system loss (shadowing, path, fading) Lp : penetration loss M: design margin (to address mobility, interference, reliability and implementation) For different terminal proles, the gains and losses and also design margins are different. Different terminal proles [1] are dened as follows. a. Fixed outdoor Located on the rooftop or outside the building walls. Usually for delay sensitive services such as VoIP also Ethernet Connection to the indoor terminal via cable High data rate is possible at a large range with low impact on the channel resources This type of terminals are mostly suited to small-tomedium enterprises (SMEs) Requires high cost of equipment and implementation b. Fixed portable indoor unit Located indoor close to the window or outer wall Unit is portable within the indoor space and has smaller coverage range Requires power supply Suitable for residential access Self-installation Lower cost compared with the xed outdoor terminals c. Nomadic mobile unit Mobile units can be used within the outdoor and indoor space Suitable for individual customers requiring particular services

3.8 Dimensioning for Coverage

91

However, customers usually prefer to have a combination of different networks. As mentioned earlier, received signal strength needs to meet at least a signal strength threshold which itself depends on SNR threshold and noise oor (Nth ). Noise oor refers to the sum of the all noise sources and SNR threshold has to have a particular level for the system to work acceptably. The SNR threshold for a received signal is typically dened to have less than 106 bit error rate when decoding at receiver [1]. However, SNR threshold varies with the modulation and coding scheme. To estimate the cell footprint we consider factors such as cell shape, network layout, number of sectors per cell to identify the operating system range. Operating system range depends on the deployment of different scenarios. If the network is deployed in a rural outdoor area the base station should be one that has footprint at its maximum range in use while if the same base station is used in urban area or for mobile users, the footprint is a percentage of the maximum footprint as overlapping is essential for handoff and a margin of about 10% is considered for this reason. In terms of cell shape, although square cells are more suitable for the xed networks, hexagonal cells are employed for implementation of the mobile network [1]. Assuming r as the operating range and F the footprint of a cell, the relationship between them is dened as: Fsq = 2r 2 3 2 3r Fhex = 2 (3.24) (3.25)

Using hexagonal cells, larger footprint are supported compared to the square cells. Therefore the number of required base stations is less than deployment by square cells. However, for achieving high capacity, using square cells is more suitable because they provide more capacity per unit area unit. For mobile network deployment hexagonal cells result in lower cost according to the number of required base stations. In some cases hybrid scenarios are in use, and deployment should be based on the worst scenario [1]. For instance if indoor and outdoor deployments are required, the planning should be based on indoor terminals because they operate in the shorter range and if the planning only meet the outdoor requirement, indoor coverage face the problem of blind

92 Planning of WiMAX and LTE Networks spots and out of coverage areas, but if plan for the indoor terminals with shorter range the required range for the outdoor terminals has already been met. The number of cells required is a function of the modulation scheme used. The more agile the modulation scheme is the higher the bit rate and the smaller the cell coverage. Hence, more base stations are required for modulation schemes which use more bits per subcarrier. BPSK uses 1 bit per subcarrier, QPSK 2 bits, 16 QAM 4 bits and 64 QAM uses 8 bits per carrier. The size of the cell is a function of the modulation type and the scheme is shown in Figure 10 Different modulation types can reach different cell range. From Figure 10, the largest cell range is achieved using BPSK 1/2. The smallest cell range is provided by 64 QAM 3/4. The cell area of coverage is inversely proportional to the capacity obtainable with a given modulation type. Thus 64 QAM 3/4 provides the largest cell capacity. The area of coverage A for each modulation type is given by the expression: 3 3 2 2 (dk dk1 ) (3.26) A= 2 Each modulation covers a percentage of the total area of the cell and provides some known capacity. Therefore the capacity of the cell can be estimated to be
M

C=
k=1

Cmodulation k

Amodulation k ATotal

(3.27)

In Figure 3.10, M = 7. 3.8.1 WiMAX Cell Capacity The use of over subscription ratio (OSR) as a criterion for design is a popular method when starting to develop the characteristics and parameters of a WiMAX cell (base station). OSR is the ratio of the total subscribers trafc demand over the reference capacity of the base station when taking into account adaptive modulation. The reference capacity of the base station is the available bit rate obtainable with the lowest modulation scheme. In the WiMAX case, this reference is provided by BPSK 1/2 which is the lowest modulation scheme served by the base station as shown in Figure 3.10. This reference capacity is given by the expression Cref = F F Tused 2TS (3.28)

3.8 Dimensioning for Coverage

93

BPSK 1/2 QPSK 1/2 QPSK 3/4 16QAM 1/2 16QAM 3/4 64QAM 2/3

d1

d2

d3 d4 d5 d6 d7

64QAM 3/4

Fig. 3.10 Size of WiMAX cell as a function of modulation type.

From Table 3.2 with TS = 40s and 256 FFT (7 MHz bandwidth and subcarrier spacing of 31.25 kHz), the reference capacity is 3.2 Mbps and this value depends on the cyclic prex value and TS . The subcarrier spacing used here is fairly large and better reference capacities can be obtained with more ne grain subcarrier spacing. The following steps could be followed in the design specications for the base station. Step 1 Dene the service area of interest. This is the area where the different types of services desired by the subscribers are provided. This means for example that the city of Sydney is taken rst and divided down into urban, high density urban, suburban, and rural areas. Taking for example the high density urban (CBD) area, repartition the area according to the types of subscribers who are

94 Planning of WiMAX and LTE Networks likely going to use services at different bit rates from this base station. For example 100 subscribers at 1 Mbps 200 subscribers at 512 kbps 300 subscribers at 256 kbps and 400 subscribers at 128 kbps

Thus this base station will handle 1000 subscribers at a time at the different bit rates specied. With this number of subscribers it is necessary to also estimate the total subscriber trafc required. Step 2: Create Sites According to the OSR For the sake of illustration, we assume that the OSR is between 1 and 2. Therefore the total trafc demand is either equal to the reference capacity or twice the reference capacity. The average trafc demand based on the subscriber distribution in Table 3.3 the average trafc demand is: Cavg = (100 1000000 + 200 512000 + 300 256000 + 400 128000)/1000 = 330.4kbps Thus at a minimum bit rate of 2,820, each base station need to be able to connect 2820/330.4 = 8 (8.54) subscribers. At the high OSR of 2, the base station can handle at most 17 subscribers. At the modulation rate of 64QAM 3/4, the base station can connect 38 subscribers. Consider a practical example where a WiMAX operator has the following service classes to attend to in the form of Platinum subscribers at a VBR of 1 Mbps minimum reserved (OSR 10), 3 Mbps maximum sustained (OSR 20), SME users, 5% of total number of subscribers
Table 3.3 Subscriber distribution. Modulation scheme QPSK 1/2 16QAM 1/2 16QAM 3/4 64QAM 2/3 64QAM 3/4 Threshold (dBm) 97 91 88 83 82 Bit rate 2, 820 5, 640 8, 545 11, 280 12, 818 C/(N + I) dB 3 9 12 17 18

3.9 Dimensioning for Capacity and Frequency Planning

95

Gold subscribers at a VBR, 540 kbps minimum reserved (OSR 10), 1 Mbps maximum sustained (OSR 20), SOHO users, 15% of total number of subscribers Silver subscribers at BE, 1 Mbps maximum sustained (OSR 20), residential users, 80% of total number of subscribers The average capacity per subscriber for this case follows the same method used earlier and is 5 1000 (3000 1000) 15 + + Cavg = 100 10 20 100 (1000 540) 85 1000 540 + + 15 20 100 20 = 61.35 kbps/subscriber At a raw bit rate of 9.6 Mbps for a base station in Table 3.2, the base station can therefore service (9.6 Mbps/61.35 kbps) 156 subscribers using QPSK with two bits per symbol. This number of subscribers would be reduced to a half (78 subscribers) if a FEC of 1/2 to three quarters (117 subscribers) if the FEC is 3/4. The range of this base station is 2.4 km as given in Table 3.2. Overheads in the PHY and MAC layers will reduce this subscriber base. Step 3: Determine the number of base stations If the total geographical area to be covered is A km square, what is the number of base stations to deploy? We assume hexagonal cells. Total Area 2.598 R 2 For a 30, 000 km square total area with base stations of radius 2.4 km, 2000 base stations are required. NBaseStations =

3.9 Dimensioning for Capacity and Frequency Planning

After accomplishing the coverage planning, capacity dimensioning and planning needs to be completed. Capacity dimensioning is performed to ensure that the capacity offered to the customers satises the required capacity according to the number of subscribers, type of service using by them and data trafc transmitted throughout the network [1]. The outcome of this process is determination of the number of the sectors per cell to address the need of required

96 Planning of WiMAX and LTE Networks

Table 3.4 Ethernet rate per modulation scheme base on 802 16e (Adapted from [1]). PHY Mode (CTC) QPSK1/2 QPSK3/4 16QAM1/2 16QAM3/4 64QAM1/2 64QAM3/4 SINR(dB) 2.9 6.3 8.6 12.7 16.9 18.0 DL(Mbps) 5 MHz 1.4 2.2 2.8 4.3 5.8 6.5 UL(Mbps) 5 MHz 0.9 1.3 1.7 2.6 3.5 3.9 DL(Mbps) 10MHz 2.7 4.3 5.8 8.6 11.6 12.9 UL(Mbps) 10 MHz 1.7 2.7 3.6 5.4 7.1 8.2

data rate for the subscribers using different services [1]. By estimation of the number of the subscribers per sector, total capacity of a cell can be identied, subsequently the capacity of whole system. To calculate the number of the subscriber per sector, the capacity of the sector needs to be identied. Average throughput of a sector is prescribes by the vendor, but it mainly depends on the deployment scenario. Different modulation schemes are used by IEEE802.16e standard. Based on the modulation scheme using in each sector, the throughput varies. Table 3.3 suggests the different modulation technique and coding and corresponding throughput in both UL and DL according to the bandwidth. According to the Table 3.3, the throughput will be maximized using 64QAM-1/2 in the 10 MHz bandwidth in the interference free environment. The equipment specication as usually explained in a table shows the modulation schemes and required SINR and Ethernet throughput for UL and DL using different bandwidth [1]. The data rate achieved through a sector is calculated using different parameters according to the modulation scheme and coding [3]: Rb = Rs Where: M: modulation gain (2 for QPSK, 4 for 16-QAM, 6 for 64-QAM) C: coding rate (1/2, 3/4, ) Rr : repetition rate of 1,2,4,6 Rb : bit rate Rs : symbol rate MC Rr (3.29)

3.9 Dimensioning for Capacity and Frequency Planning

97

Although different modulation schemes are used in different sectors, because different terminals experience divergent interference due to frequency reuse, IEEE802.16, recommends a default assumption for the average sector throughput. Based on this assumption, the average sector throughput is dened as the throughput corresponding to the16-QAM modulation scheme and 1/2 coding rate; however, the throughput will be higher if deployment conditions are desirable. Since the average sector throughput is determined, customer data should be examined in terms of type of services, average data rate and VoIP Committed Information Ratio. After that a graph is required to show the subscribers per sector versus throughput. The next step is to determine the number of the required sectors identied by dividing the number of the customers per area with the number of customers per sector. Assume the number of customers per unit area is S, the uplink bit rate to be Rb,U L and the number of customers per sector is SS , then the number of sectors is NS = S SS (3.30)

The bit rate per sector can be determined based on system parameters and is given by the expression: NdataDL 1 overhead bits per symbol x xTDDdown/upratio NFFT 1 + Tg (3.31) The TDD down/up ratio between the UL and DL times and may be 3:1 and the overhead time includes times used for synchronisation, initialisation and headers. However, adaptive modulation is one of the important factors that affect the capacity of the network [1]. As described earlier , WiMAX is able to adopt different modulation schemes for data transmission based on the link quality. For example if the SINR is high, transmission can be done using higher order of the modulation schemes such as 64-QAM to achieve higher throughput while in the case of weak SINR lower orders of modulation schemes such as QPSK can be employed. Alongside the adoption of different modulation schemes, to increase the capacity of the network and minimize the interference between the co-channel Rb,DL = B.n.

98 Planning of WiMAX and LTE Networks cells, WiMAX uses powerful sub-channelization modes which are PUSC (Partial Usage of Sub-Channels) and FUSC (Full Usage of Sub-Channels) [1, 4, 5]. FUSC refers to the usage of all sub-carriers within a cell, while PUSC is dened as allocation of specic group of sub-carriers to each sector within a cell, which there are typically 3 sectors per cell. Using PUSC mode allows the network to use the frequency reuse factor of one when the spectrum is limited without interference between adjacent sectors, because they use divergent subcarriers. This scheme is very efcient as there is no need for guard band also provides more reliability of links in both uplink and downlink by using different sub-carriers in the neighbouring cells subsequently reducing the chance of interference [1, 4]. Whereas FUSC that is used only in downlink PUSC can be used in both direction of uplink and downlink. The reason for that FUSC can not be used in uplink is because of the nature of uplink which is more unpredictable than downlink [5]. To plan the frequency channels to achieve higher capacity and optimum use of the channels, there are some factors that need to be considered in either of uplink and downlink. Planning for downlink is easier in comparison with uplink because the major source of interference in downlink is the neighbouring base station using the same frequency channel; therefore it is possible to calculate the SINR in each spot and prepare an interference matrix for each base station in the service area and the proper modulation scheme for each spot [5]. For example in a location has suitable condition higher modulation order can be used to enhance the capacity of the network. However, most of the downlink interference is expected at the borders of cells where the SINR is low and signal is more likely to collide with other signals on the same carrier frequency [5]. To overcome this problem, WiMAX divides the TDD frame, which consists of two portions of uplink and downlink, each includes some slots and each slot in turn consists of 48 sub-carriers, to two parts in downlink [5]. At the rst part of the downlink frame, base station adopt PUSC mode to transmit data to the subscribers located at the edge of the cell to reduce the possible interference while using FUSC mode in the second part of the downlink frame, to transmit data to the subscribers near to the base station with higher SINR for increasing the capacity and nally efcient use of the spectrum [5]. In spite of downlink, uplink planning is more complicated as it is less predictable. The subscriber station can be any where within the service area also any where within a particular cell therefore it is difcult to determine

3.9 Dimensioning for Capacity and Frequency Planning

99

the receiving level of SINR consequently the level of interference. Interference can vary not only by changing the location of the user but also by the number of the slot that it uses to transmit the data on. Using more slots by a subscriber leads to more number of sub-carriers and subscribers it collides with [5]. According to all those reasons mentioned above, the uplink can use only PUSC scheme for data transmission to achieve a robust network. To allocate an optimum modulation scheme to the link, base station estimates the SINR according to the feedback that it receives from the subscriber station through the previous received frame. Since in mobile WiMAX system, the user changes the channel frequently because of mobility, the base station needs to be aware of the current channel that user is using; therefore to allocate the best possible modulation scheme to the connection, the base station has to use any of the average SINR received previously, the best SINR or the worst SINR. The average SINR is more suitable since the selection of the worst SINR results in capacity reduction due to data transmission on lower that optimal modulation order, while assigning the modulation scheme based on the best amount of SINR causes the less robust links due to high error occurrence and data retransmission [1, 5]. What is done on capacity dimensioning within the access network will assist to determine the required capacity for the backhaul network. It is discussed in some detail in the next section. 3.9.1 Backhaul Dimensioning The backhaul network links a number of base stations and ASN-gateways, via a radio interface or bre rings [1]. Base stations can be connected together through a point to point, if the number of the required base stations are high, or point to multipoint radio interface, if the base stations are close to each other. Also the base stations need to be connected to a gateway within the backhaul network using an IP transport or IP link, which carries the IP packet between the base station and the gateway. To evaluate or plan a backhaul network, the information rate of the backhaul, the number of the base station and gateways also the position of them needs to be determined. The backhaul information rate increases with the number of the sectors per a cell, also the sectors throughput in a linear manner. However, the sectors throughput varies with the deployment scenarios. Total TDD rate that can be

100 Planning of WiMAX and LTE Networks delivered by the backhaul is the product of the number of the sectors and each sectors throughput. To calculate the FDD rate, the total throughput needs to be multiply by DL/(DL+UL) rate [1]. In accord with WiMAX forum, using MIMO system and a channel bandwidth of 10 MHz, the uplink air interface can support 28 Mbps per sector peak data rate and 63 Mbps at downlink [1]. Employing three sectors per a base station, total peak data rate that a base station can deliver in uplink and downlink is 84Mbps and 189 Mbps respectively [1]. To calculate the total data rate supported by a base station, two other factors need to be determined, which are the multiplexing technique in used and the uplink and downlink rate. Assuming FDD technique and 3:1 downlink and uplink rate, the peak downlink data rate is 46 Mbps/sector and 8 Mbps/sector for uplink, which leads to total peak data rate of 46 Mbps per sector for each base station, as the total data rate in a FDD system is equivalent to the maximum rate between uplink and downlink [1]. Therefore by multiplying the total throughput for each sector by three, total peak data rate per a base station is calculated, which is approximately 150 Mbps in this case. It means that the link between the base station and gateway needs to be able to have the capacity equal to 150 Mbps [1]. In contrast, if TDD technique is in used in the same scenario, the peak data rate per each sector in uplink and downlink is 1.83 Mbps and 13.60 Mbps respectively which results in total data rate of about 16 Mbps per sector, as the sum of uplink and downlink data rate in TDD systems result in total data rate [4]. Triple the throughput of each sector result in peak data rate of 48 Mbps per a base station. Choosing either FDD or TDD, also the downlink and uplink rate depends on the deployment scenario and the type of application is due to be delivered in the network. However a comparison between TDD and FDD assist to select the best technique that suites the providers. Different metrics can be considered for the evaluation of the backhaul planning and the techniques that need to be deployed for achieving the better performance, such as different service type to be offered, the available bandwidth, data rate, complexity and cost also the number of subscribers that can be served by the base station. In terms of bandwidth availability, if the frequency band is limited, TDD is more likely to have the better performance than FDD; since FDD needs different frequencies for downlink and uplink, while TDD send the uplink and

3.9 Dimensioning for Capacity and Frequency Planning

101

downlink data trafc on the same frequency band but different time slots. However, planing for different service types, FDD is more suitable for the applications which require more bandwidth such as video, whereas TDD which is proper for application such as voice that requires lower bandwidth. Furthermore, FDD provides higher data rate than TDD but TDD is far less complex therefore lower deployment cost [4]. However, operators can benet the high number of subscribers supported by FDD technique when they demand applications with lower bandwidth requirements. To achieve an efcient backhaul dimensioning, it is recommended by [1] to dimension the backhaul network and anticipate the total required capacity of the WiMAX backhaul network based on the predicted demand for service by dividing the total available spectrum into three different parts, each part serving a portion of the users with the same required quality of service applications. It is suggested in [1] that this method of dimensioning requires less backhaul deployment cost initially and scale well if the number of subscribers increases by adding more backhaul. To deploy that, assuming three different types of applications are in demand dened as video calls with required high QoS guaranteed, voice calls which requires less QoS guaranteed compare to video calls and nally best effort service, which provides services such as the Internet access with no guarantee of QoS requirement. In this case, the aim is to identify the total capacity needed for the backhaul link between two base stations or between the base station and gateway to support all different service applications efciently. The rst step is to divide the applications into three groups known as EF class (Expedited Forwarding), AF class (Assured Forwarding) and BE class (Best Effort) [1]. The applications such as video calls are placed in the EF category, which require absolute amount of delay and very sensitive to delay. AF class is include the applications with average delay and less sensitive to delay compare to the EF class applications, such as voice calls. BE class consists of the applications with no sensitivity to delay. The next step is to determine how many subscribers are using the spectrum simultaneously at each class, which is assumed to be N voice calls, M video calls and few BE type applications to be served by a base station. Based on the assumption that a portion of the spectrum is used by n EF subclasses and other portion of the spectrum is divided to be used by m AF subclasses and few BE, also using a

102 Planning of WiMAX and LTE Networks weighted fair queuing scheduler with h as a vector dening the weights used in the scheduling to determine the distribution of the spectrum for the AF and BE class, the capacity requires by different classes are calculated by [1]: ciEF = ciAF = cBE = Where: ciEF : each EF subclass needs to have at least this amount of capacity ciAF : total bandwidth available for AF class cBE : total bandwidth available for BE class c: capacity of the backhaul link hEF : weight for EF class hAF : weight for AF class hBE : weight for BE class hEF c i AF + +h hAF c i + hAF + hBF c i + hAF +
n EF i=1 hi

hBE hBE hBE

(3.32) (3.33) (3.34)

n EF i=1 hi

n EF i=1 hi

Now the backhaul link capacity can be determined by accumulating the data result from the calculation of capacity needed for each subclass.

References
[1] Yan Zhang. WiMAX Network Planning and Optimization. USA: CRC Press, 2009. [2] Johnson I Agbinya. IP Communications and Services for NGN. New York: Taylor and Francis, 2009. [3] Bharathi Upase, Mythri Hunukumbure, Sunil Vadgama. Radio Network Dimentioning and Planning for WiMAX Networks. 43.4, pp. 435450 Available: http://www.fujitsu.com/downloads/MAG/vol43-4/paper09.pdf. [4] WiMAX Forum. Mobile WiMAX-Part I: A Technical Overview and Performance Evaluation. WiMAX Forum, pp. 153, 2006. [5] Kai Dietze Ph.D, Ted Hicks, WiMAX Uplink and Downlink Design Consideration. Internet: http://www.edx.com/les/wimax_paper_2v4-1.pdf, [Sep.19, 2009]. [6] Steve Wisniewski. Wireless and Cellular Networks. USA: Pearson Prentice Hall, 2005. [7] Katzela, I. and M. Naghshineh, Channel Assignment Schemes for Cellular Mobile Telecommunication Systems: A Comprehensive Survey, IEEE Personal Communications, pp. 1031, June 1996.

References

103

[8] Jeffrey G. Andrews, Arunabha Ghosh, Rias Muhamed. Fundamentals of WiMAX. USA: Prentice Hall, 2007. [9] Rhode and Schwarz, WiMAX General Information about the standard 802.16 (application note), 2006, pp. 134. [10] WiMAX Capacity White Paper, SR Telecom Inc., Canada, 2006, pp. 1-34.