You are on page 1of 13

Journal of the Chinese Institute of Engineers, Vol. 32, No. 5, pp.

585-597 (2009) 585

CELL PLANNING AND CHANNEL THROUGHPUT OF MOBILE


WiMAX AT 2.5 GHz

Jeich Mar * , Chin-Chung Ko, Chung-Haw Li, and Shao-En Chen

ABSTRACT
A combination of COST-231 Hata and SUI Erceg models is presented to predict
the propagation path loss of 2.5 GHz Mobile WiMAX in urban, suburban, flat-terrain
with light tree (rural A), flat-terrain with heavy tree (rural B) and hilly-terrain rural
(rural C) environments. The cell sizes for five different terrain areas, three antenna
modes and 90% service reliability are estimated through the downlink link budget
analysis. Based on the radio coverage calculation results, the call holding time in a
given cell and channel throughput of the Mobile WiMAX are simulated to observe the
relationship among mobile speed, antenna mode, operation environment and channel
throughput. The effect of user mobility on the handoff rate is considered in the
simulations. An example of cell planning for 2.5 GHz Mobile WiMAX is carried out
for a zone near Taipei city.

Key Words: cell planning, handoff rate, cell residence time, channel throughput.

I. INTRODUCTION modulation schemes, namely BPSK, QPSK, 16-QAM


and 64-QAM, can be selected for each user to guaran-
Mobile WiMAX is a broadband wireless solu- tee the target BER. Mobile WiMAX supports a full
tion that enables the convergence of mobile and fixed range of smart antenna technologies to enhance both
broadband networks through a common wide area broad- coverage and channel throughputs (WiMAX Forum,
band radio access technology and flexible network 2007). The 1 × 2 SIMO is a typical Mobile WiMAX
architecture. The Mobile WiMAX Air Interface adopts baseline BS antenna mode. A single transmission an-
Orthogonal Frequency Division Multiple Access tenna is used at each end of the link, in which the re-
(OFDMA) technology for improving multi-path per- ceived signal is enhanced through the use of diversity
formance and supporting non line-of-sight (NLOS) in the dual receiver antennas at both the BS and the
environments operation in the 2-11 GHz band. This mobile station (MS). The second base station antenna
year, the 2.5 GHz frequency carrier band has been re- mode provides a 2 × 2 multiple input multiple output
leased for the use of service providers in Taiwan. Cell (MIMO) configuration, which offers space-time cod-
planning could aid service providers to optimize base ing (STC) and spatial multiplexing (SM) modes. With
station (BS) configurations, antenna mode selections, STC antenna mode, identical downlink data streams
and site locations, in order to meet quality of service, are sent from each transmission antenna providing space
system capacity and service requirement standards at and time diversity. In an environment with rapid
minimum cost. The Mobile WiMAX specification multipath fading, STC enhances the signal-to-noise ratio
proposes a 1024 orthogonal frequency division multi- (SNR) of the received signal at the mobile station to
plex (OFDM) approach to cope with expected chan- enable support of higher modulation efficiency bursts,
nel impairments (Liu and Li, 2005). Various and thus enhances the downlink capacity as well as
downlink range. With SM antenna mode, each of the
base station transmission antennas sends a different
*Corresponding author. (Tel: 886-3-4638800 ext. 7320; Fax: downlink data stream. This technique uses multiple
886-3-4554264; Email: eejmar@saturn.yzu.edu.tw)
The authors are with the Department of Communications
paths to distinguish between the different data streams
Engineering, Yuan-Ze University, 135 Yuan-Tung Road, Jungli, and has the theoretical potential to double the down-
Taoyuan 320, Taiwan, R.O.C. link capacity under favorable channel conditions.
586 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

The coverage of WiMAX with 256 OFDM at 450 from (Abhayawardhana et al., 2005) (Erceg and
MHz and 3.5 GHz are analyzed and compared (Javornik Greenstein, 1999). The pass loss and shadow fading
et al., 2006) using the Erceg path loss model for three are taken into account in the models.
different environments. In this paper, the cell size of
2.5 GHz Mobile WiMAX is estimated from the link 1. COST-231 Hata Model
budget using the COST-231 Hata propagation model
(Javornik et al., 2006) and the Stanford University The basic pass loss equation in dB for the COST-
Interim (SUI) Erceg model (Abhayawardhana et al., 231 Hata model is (Abhayawardhana et al., 2005)
2005) (Erceg and Greenstein, 1999). The channel
throughput of Mobile WiMAX has been simulated by PL = 46.3 + 33.9log10( f ) – 13.82log 10(h b) – ah m
presenting multimedia call services and cell layout
+ (44.9 – 6.55log 10(h b))log 10(d) + c m, (1)
scenarios to evaluate the channel throughput. To
determine the channel throughput in terms of aver-
where f is the frequency in MHz, d is the distance
age call holding time in a given cell, we need to com-
between receiver and transmitter in km, and h b is the
pute the cell residence time and handoff rate for new
BS antenna height above ground in meters. The pa-
generation and handoff calls at different MS speeds.
rameter c m is defined as 0 dB for suburban or rural
The effect of 2 × 2 MIMO antenna modes and ve-
environments and 3 dB for urban environments. The
hicle speed on cell size and channel throughput are
ah m is defined for urban environment as
also analyzed and calculated.
The rest of the paper is organized as follows.
ah m = 3.20(log 10(11.75h r)) – 4.97,
In Section II a combination of the COST-231 Hata
model and the SUI Erceg model is proposed for cell for f > 400 MHz (2)
size estimation in five different areas; Section III
determines the fade margin and the coverage prob- and for suburban or flat rural environments,
ability under 90% service reliability conditions; the
radio coverage calculation results and a cell planning ah m = (1.1log 10 f – 0.7)h r – (1.56log 10 f – 0.8),
example are shown in Section IV; the simulations of
(3)
cell residence time and handoff rate at different MS
speeds in a given cell together with channel through- where h r is the MS antenna height above ground in
put of Mobile WiMAX are performed in Section V. meters.
Section VI concludes this paper.
2. Stanford University Interim (SUI) Erceg Model
II. CELL SIZE ESTIMATION IN FIVE
DIFFERENT AREAS The SUI Erceg model is divided into three types
of terrains, A, B and C. Type A is hilly terrain with
The propagation model plays a significant role moderate to heavy tree density, representing rural
in cell planning. The COST-231 Hata model was environments and is associated with the highest path
designed at a frequency band range from 500 MHz to loss. Type B is characterized by either a mostly flat
2 GHz to be widely used to estimate path loss in mo- terrain with moderate to heavy tree density or a hilly
bile wireless systems for urban, suburban and flat terrain with light tree density. Type C is a flat ter-
rural environments. The SUI Erceg model was ini- rain with light tree density and is associated with the
tially designed for suburban areas and for a frequency lowest path loss for rural environments. The basic
band of 1.9 GHz. Refer to (WiMAX Forum, 2006), path loss equation with correction factors is presented
the COST 231- Hata model is applied for the propa- as (Abhayawardhana et al., 2005),
gation model of Mobile WiMAX in WiMAX Forum
2006. The COST 231- Hata model is based on em- PL = H + 10γ log 10( d ) + X f + X h + s for d > d 0 ,
d0
pirical results in the 2 GHz band and tends to make (4)
very conservative prediction for 2.5 GHz. The SUI
Erceg model is another model commonly used for the where d is the distance between the receiver and trans-
2.5 GHz band and predicts cell sizes that are greater mitter in meters, d 0 = 100 m and H is defined as
than those of real cells. Based on the prediction
4π d 0
results, the operator will need to perform model cali- H = 20log 10( ), (5)
λ
bration through field tests for the final cell planning.
Both models also can use correction factors to be where λ is the wavelength. The parameter γ is a
widely applied to other frequency bands (Javornik et Gaussian random variable over the population of
al., 2006). The formulas for both models are taken macrocells within each terrain category. It can be
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 587

Table 1 The parameters of the SUI Erceg model 170


(Abhayawardhana et al., 2005)
160
Terrain Type Type Type
parameter A B C 150

e 4.6 4.0 3.6 140


g(m –1 )

Path Loss (dB)


0.0075 0.0065 0.005
130 Erceg C
k(m) 12.6 17.1 20.0 Erceg B
σγ 0.57 0.75 0.59 Erceg A
120
µσ 10.6 9.6 8.2 COST-231 suburban
COST-231 urban
σσ 2.3 3.0 1.6 110

100

90
written as (Erceg and Greenstein, 1999)
80
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
γ = (e – ghb + k/h b) + x σ γ , (6)
Cell size (km)

where h b is the height of the base station antenna in Fig. 1 Path loss curves of Mobile WiMAX using the COST-231
meters, σ γ is the standard deviation of γ , x is a zero- Hata model and the SUI Erceg model
mean Gaussian variable of unity standard deviation
N[0, 1], and e, g, k and σ γ are all data-derived con- WiMAX are (WiMAX Forum, 2006)
stants for each terrain category. The shadow fading
component s varies randomly from one terminal lo- • Frequency = 2.5 GHz
cation to another within any given macro-cell. It is a • MS height = 1.5 meter
zero-mean Gaussian variable and can be expressed • BS height = 32 meter
as (Erceg and Greenstein, 1999)
Fig. 1 compares the propagation path loss (PL) cal-
s = yσ, (7) culated by the COST-231 Hata model and the SUI
model for urban, suburban and rural environments.
σ = µσ + zσσ , (8) It is interesting to reveal that the path loss obtained
by the COST-231 Hata model is higher than that of
where y and z are the zero-mean Gaussian variables the SUI Erceg model, given the same cell size. So
of unit standard deviation N[0, 1], σ is the standard we may use the COST-231 Hata model to estimate
deviation of s, µσ is the mean of σ, and σσ is the stan- the cell size of Mobile WiMAX for urban and subur-
dard deviation of σ . µ σ and σ σ are both data-derived ban environments, and use the SUI Erceg model for
constants for each terrain category. The numerical three different terrain categories of rural environments.
values of the above parameters are given in Table 1.
The correction factors of the model for the op- III. LOGNORMAL FADE MARGIN AND
erating frequency and for the MS antenna height are COVERAGE PROBABILITY
(Abhayawardhana et al., 2005)
In the link budget analysis, fade margin data for
f
X f = 6.0log 10( ), urban, suburban and rural environments are needed,
2000
together with the coverage probability and service
h
X h = –10.8log 10( r ) for Terrain types A and B reliability. Coverage probability is the probability at
2
location l in a cell with radius L, that the received
hr
= –20.0log 10( ) for Terrain type C, (9) signal power is above threshold.
2
where f is the frequency in MHz and h r is the height 1. Coverage Probability
of the MS antenna above ground in meters.
It is assumed that the local mean signal strength
3. Cell Size Estimation in an area at a fixed radius from a particular base
station antenna is log-normally distributed. For log-
The COST-231 Hata model and the SUI Erceg normal fading environments, we define the received

model are numerically calculated to obtain the signal strength W(dBm) with mean W and standard
relationship between cell size and path loss, as shown deviation σW. The probability density of W is (Jakes,
in Fig. 1, in which the parameters of the Mobile 1993)
588 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

1 e – 12 ( Wσ– W ) 2 . Table 2 Lognormal fade margin coverage prob-


p(W) = (10)
2π σ W W
ability
σW α P W0(L) F.M
The probability that W exceeds the threshold W 0 is
∞ Urban 6 4 70% 3.2
Suburban 8 3.4 76% 5.56
PW0(l) = P[W ≥ W0] = p(W)dW
Rural 12 2.5 82% 11.2
W0

W0 – K + 10αlog( l ) (11)
= 1 – 1 erf ( L ),
2 2 2σW modulation and coding (AMC) and two basic types
u of uplink sub-channels including PUSC, and AMC.
where the error function is erf(u) = 2π exp(–z 2)dz, In the PUSC mode, among all sub-carriers, only 720
0
K-W 0 is defined as the fade margin in dB at l = L to of them carry information. One hundred and twenty
assure reliable operation of the communication link, sub-carriers transmit the pilot tones. In the FUSC
K is the mean received signal strength in dB at cell mode, among all sub-carriers only 768 of them carry
boundary l = L and W0 is the threshold of the receiver. information. Eighty-three sub-carriers transmit the
We assume that the behavior of the mean value of pilot tones (Liu and Li, 2005).
signal strength follows an l –α law, thus Six cases of downlink link budget analysis are
summarized in Table 3(a) for Mobile WiMAX in the
W = K – 10αlog( l ) , 2 ≤ α ≤ 4. (12) suburban environment (WIMAX Forum, 2006), case
L 1 is a basic type of PUSC downlink sub-channel using
2. Service Reliability 2 × 2 MIMO STC antenna mode; case 1 * is a basic
type of PUSC downlink sub-channel using 2 × 2 MIMO
Service reliability F u determines the proportion SM antenna mode; case 2 is the PUSC downlink sub-
of useful service cell area within a circle of radius L channel using 1 × 2 SIMO antenna mode; case 3 is the
for which the signal strength received by an MS ex- FUSC downlink sub-channels using 2 × 2 MIMO STC
ceeds a given threshold W 0. If P W0 is the probability antenna mode; case 3* is the FUSC downlink sub-chan-
that the received signal W exceeds W 0 in an incre- nels using 2 × 2 MIMO SM antenna mode; case 4 is
mental area dA (Jakes, 1993), then the FUSC downlink sub-channels using 1 × 2 SIMO
antenna mode. The differences among the six cases
Fu = 1 2 PW0dA . (13) are antenna modes and downlink sub-channel types.
πL Finally the maximum allowable path loss results (in
Letting dB) for the six cases in suburban areas are calculated
and listed in Table 3(a), where
W0 – K + 10αlog( l )
m= L ,
2σW • EIRP = Tx Power per Antenna Element + Cyclic
Combining Gain + Tx Antenna Gain + Pilot Power
10αloge (14) Boosting Gain
n= ,
2σW • Rx Sensitivity = Rx Noise Figure + Thermal Noise
we get + SNR Required
• System Gain = EIRP + Rx Antenna Gain + Rx An-
Fu = 1 {1 – erf (m) + exp( 1 – 2mn )[1 – erf ( 1 –nmn )} . tenna Diversity Gain – Rx Sensitivity
2 n2 • Maximum Allowable Path Loss = System Gain –
(15)
Total Margin
We use the above formulas to calculate lognormal
fade margin and coverage probability at the cell edge Table 3(b) shows the cell sizes for the six cases in
for a Mobile WiMAX system, as shown in Table 2 five environments, namely urban, suburban, flat-ter-
(Parsons, 1992; Rappaport, 2002), where the service rain with light trees (rural A), flat-terrain with heavy
reliability is set as 90%. trees (rural B) and hilly-terrain rural (rural C)
environments, respectively. Using lognormal fade
IV. RADIO COVERAGE CALCULATION margin and interference margin values specified in
RESULTS (Harri and Antti, 2000) for urban, suburban and rural
environments, the maximum allowable path losses for
There are three basic types of downlink PUSC and FUSC types downlink sub-channels of
sub-channels including fully used sub-carrier (FUSC), Mobile WiMAX with 2 × 2 MIMO STC and SM an-
partially used sub-carrier (PUSC), and adaptive tenna modes and 1 × 2 SIMO antenna mode are
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 589

Table 3 (a) Downlink link budget for Mobile WiMAX in suburban area. (b) Cell size of Mobile WiMAX
for different environments. ( ) * means the 2 × 2 MIMO SM antenna mode
(a) (b)
Base Station infrastructure Case 1(1*) Case 2 Case 3(3*) Case 4 Units Urban Case 1 Case 2 Case 3 Case 4 Units
Tx Power per Antenna 10.0 10.0 10.0 10.0 Watts Log Normal Fade Margin 3.2 3.2 3.2 3.2 dB
Element
Fast Fading Margin 6 6 6 6 dB
Number of Tx Antenna 2 1 2 1
Elements Interference Margin 3 3 3 3 dB
MIMO Gain (2 Antennas) 3.0(0)* 0 3.0(0)* 0 dB Penetration Loss 10 10 10 10 dB
Tx Antenna Gain 15.0 15.0 15.0 15.0 dBi Total Margin 22.2 22.2 22.2 22.2 dB
Pilot Power Boosting Gain -0.7 -0.7 -0.7 -0.7 dB Maximum Allowable Path 137.8 134.8 137.7 134.7 dB
EIRP 57.3 54.3 57.3 54.3 dBm Loss (131.8)* (131.7)*
Sub-channel type PUSC PUSC FUSC FUSC Log Normal Shadowing SD 6 6 6 6 dB
Number of Occupied Sub- 840 840 851 851 Cell size (COST 231 Urban) 0.6791 0.5576 0.6746 0.5539 km
Carriers (0.4578)* (0.4548)*
Power per Occupied Sub- 28.1 25.1 28 25 dBm Suburban
Carrier Log Normal Fade Margin 5.56 5.56 5.56 5.56 dB
Mobile Unit (Handset Indoor) Fast Fading Margin 6 6 6 6 dB
Rx Antenna Gain -1.0 -1.0 -1.0 -1.0 dBi
Interference Margin 2 2 2 2 dB
Rx Antenna Diversuty Gain 3.0(0)* 3 3.0(0)* 3 dB dB
(2 Antennas) Penetration Loss 10 10 10 10
Rx Noise Figure 7.0 7.0 7.0 7.0 dB Total Margin 23.56 23.56 23.56 23.56 dB
Margins Maximum Allowable Path 136.4 133.4 136.3 133.3 dB
Loss (130.4)* (130.3)*
Log Normal Fade Margin 5.56 5.56 5.56 5.56 dB
Log Normal Shadowing SD 8 8 8 8 dB
Fast Fading Margin 6.0 6.0 6.0 6.0 dB
Cell size (COST 231 0.7572 0.6217 0.7522 0.6176 km
Interference Margin 2.0 2.0 2.0 2.0 dB Suburban) (0.5105)* (0.5071)*
Penetration Loss 10.0 10.0 10.0 10.0 dB
Rural
Total Margin 23.56 23.56 23.56 23.56 dB
Log Normal Fade Margin 11.2 11.2 11.2 11.2 dB
Mobile Rx Sensitivity
Fast Fading Margin 6 6 6 6 dB
Thermal Noise -174 -174 -174 -174 dBm/Hz
Interference Margin 1 1 1 1 dB
Sub-Carrier Spacing 10.94 10.94 10.94 10.94 kHz
Penetration Loss 10 10 10 10 dB
Modulation (coding rate) QPSK QPSK QPSK QPSK
(1/8) (1/8) (1/8) (1/8) Total Margin 28.2 28.2 28.2 28.2 dB
SNR Required -3.31 -3.31 -3.31 -3.31 dB Maximum Allowable Path 131.8 128.8 131.7 128.7 dB
Loss (125.8)* (125.7)*
Delta from Limiting cell range 0.82 0.82 0.82 0.82
distance Log Normal Shadowing SD 12 12 12 12 dB
Rx Sensitivity (per sub-carrier) -129.9 -129.9 -129.9 -129.9 dBm Cell size (Erceg model A) 1.095 0.952 1.09 0.948 km
Rx Sensitivity (composite) -100.7 -100.7 -100.6 -100.6 dBm (0.827)* (0.823)*
System Gain 160.0 157.0 159.9 156.9 dB Cell size (Erceg model B) 1.384 1.193 1.38 1.19 km
(154.0)* (153.9)* (1.01)* (1.008)*
Maximum Allowable Path 136.4 133.4 136.3 133.3 dB Cell size (Erceg model C) 1.53 1.393 1.52 1.39 km
Loss (130.4)* (130.3)* (1.097)* (1.094)*

calculated. It is noted that case 1 generates the larg- The cells planning for a zone near Taipei city
est path loss, given the same environment, so its cell are shown in Figs. 2(a), (b) and (c) for FUSC 2.5 GHz
size is larger than other cases. A base station installed Mobile WiMAX with 2 × 2 MIMO STC antenna mode,
in a rural C environment needs a plan incorporating 1 × 2 SIMO antenna mode and 2 × 2 MIMO SM an-
the largest cell sizes for all cases. The maximum al- tenna mode, respectively. The circles a and b on the
lowable path losses of PUSC 2 × 2 MIMO antenna terrain map are the cell sizes in urban and suburban
modes are almost equal to those of FUSC 2 × 2 MIMO areas, respectively, calculated by the COST-231 Hata
antenna modes. Because of having MIMO antenna model; the circles c, d and e in rural areas A, B and C
gain, the maximum allowable path losses of 2 × 2 are the cell sizes calculated by the SUI Erceg model,
MIMO STC antenna modes of both PUSC and FUSC respectively.
type are 3 dB larger than the 1 × 2 SIMO antenna The cell size in Fig. 2(c) is assumed to be identi-
mode and 6 dB larger than the single input single out- cal for all five operation environments. The ratios of
put (SISO) antenna so that the cell size of Mobile cell sizes in Figs. 2(a) and (b) to cell size in Fig. 2(c)
WiMAX using 2 × 2 MIMO STC antenna mode is are listed in Table 4, where circle a represents an ur-
larger than when using 1 × 2 SIMO antenna mode. ban area, circle b represents a suburban area, and circles
Given 90% service reliability over the entire cell c, d, and e represent rural areas A, B and C, respectively.
area, the value of 3.2 dB used for the lognormal fade It is noted that the sizes of all circles in Fig. 2(a) are
margin in the urban environment assures a 70% cov- larger than those in Figs. 2(b) and 2(c). For this reason,
erage probability at the cell edge; the value of 5.56 the Mobile WiMAX using 1 × 2 SIMO antenna mode
dB used for the lognormal fade margin in the subur- or 2 × 2 MIMO SM antenna mode needs more BSs to
ban environment assures a 76% coverage probability cover the same service area. Nevertheless, the cell
at the cell edge; the value of 11.2 dB used for the log size of the Mobile WiMAX using a 2 × 2 MIMO SM
normal fade margin in the urban environment assures antenna mode, which is identical to a SISO antenna,
a 82% coverage probability at the cell edge. is the smallest but it can greatly improve the channel
590 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

Table 4 Comparison of cell sizes in Fig. 2


ratio a b c d e
Fig 2a 2.2 2.2 1.75 1.87 1.93
Fig 2b 1.483 1.483 1.33 1.39 1.61
Fig 2c 1 1 1 1 1

(a) (b) (c)

Fig. 2 Cell planning for a zone near Taipei city for 2.5 GHz FUSC Mobile WiMAX: (a) with 2 × 2 MIMO STC antenna mode; (b) with 1
× 2 SIMO antenna mode; (c) with 2 × 2 MIMO SM antenna mode

throughput. So taking best advantage of STC and SM • Multi-data rate call ratio = D A,UGS : D A,BE : D A,nrtPS
Mobile WiMAX also supports adaptive MIMO : D A,rtPS = 1 : 1 : 1 : 1
switching. This enables dynamic switching between • System capacity = 40 equivalent slots (1 equivalent
STC and SM depending on the existing channel con- slot = 64 Kbps)
ditions at any given time (WIMAX Forum, 2007). • Call completing ratio = 0.9
• Minimum handoff to adjacent cell ratio (handoff
V. CHANNEL THROUGHPUT rate) for new generation call Rrn = 0.1
• Call management method: Complete sharing (CS)
Based on the cell planning results, the channel scheme without channel reservation (Mar and
throughputs of the Mobile WiMAX are simulated to Huang, 2000)
observe the relationship among mobile speed, antenna
mode, operation environment and channel throughput. DOCSIS 1.1 (Data Over Cable Service Interface
The digital simulation method in (Reilly, 1999) (Nelson, Specification) is adopted to define multimedia call
1995) is used to evaluate the channel throughput. The services: Unsolicited Grant Service (UGS), Best Ef-
cell layouts are given in Figs. 2(a), (b) and (c), where fort (BE), Non-Real-Time Polling Service (nrtPS) and
each antenna mode includes five different operation Real-Time Polling Service (rtPS)(IEEE Computer
environments. Different shadowing fade margin values Society and IEEE Microwave Theory and Techniques
are considered in the link-budget calculation of do- Society, 2006). UGS, BE, nrtPS, rtPS are assigned
ing cell planning. The Mobile WiMAX cluster is with D A,UGS, D A,BE, D A,nrtPS, D A,rtPS equivalent slots,
composed of concatenated segments where each seg- respectively. We assume that the data transmission
ment has installed a Mobile WiMAX system. The rate of one channel is equivalent to 64 Kbps per time
simulation parameters are listed in the following (Mar slot for Mobile WiMAX systems (VoIP application
and Huang, 2000; Mar et al., 2007): requires 64 Kbps) (Mar et al., 2007). Therefore, one
equivalent slot is required for 1x data transmission
• New call origination is modeled with Poisson dis- rate (64 Kbps) service (D A,UGS, D A,BE ), two equiva-
tribution function. Mean call arrival rate = 0.18 lent slots are required for 2x data transmission rate
calls/sec (128 Kbps) service (D A,nrtPS ), and three equivalent
• Average speed of users = 50, 100 and 120 km/hr slots are required for 3x data transmission rate (192
• System traffic load ratio = 1 Kbps) service (D A,rtPS). All types of calls have equal
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 591

priority to be serviced in a CS scheme. The call hold- time that the call will remain in progress if it experi-
ing time is defined as the time duration between the ences no forced termination due to handoff failure.
instant that the slot is occupied by a call and the in- The unencumbered call duration is a random variable
stant it is released by either call completion or call with a negative exponential probability density func-
handoff. The call holding time for a new call in sys- tion (pdf). The mean unencumbered call duration is

tem i, THn, i is the minimum of the unencumbered call T u = u –1 , in which u is the mean call unencumbered
duration T u and the cell residence time for a new call rate or call completion rate (Mar et al., 2007). The
in system i. Also the slot holding time T Hh, i for a cell residence time is defined as the time duration that
handoff call in system i is either the unencumbered an MS resides in a cell. The distance between the
call duration T u or the cell residence time T h, i for a point where a new call is originated and the point on
handoff call in system i, whichever is less. The ran- the cell boundary where an MS exits from cell is Z i.
dom variables for T Hn, i and T Hh, i are given by Then the cell residence time for a new call T n, i is
expressed as T n, i = Z i /V i. This is a function of sys-
T Hn, i = min (T u, T n, i), tem parameters such as cell size, speed and direction
of movement by the MS. The pdf of cell residence
T Hh, i = min (T u, T h, i). (16)
time T n, i for a new call in system i is derived in Ap-
pendix A.
The unencumbered call duration T u is the amount of

V i max + V i min
πri , for t = 0

2 3 2 3
8r i tV i min tV i max 2r i
fTn, i(t) = 1– – 1– , for 0 ≤ t ≤ , (17)
3(V i max – V i min) π t 2 2r i 2r i V i max

2 3
8r i tV i min 2r i 2r i
1– , for ≤t≤
3(V i max – V i min) π t 2 2r i V i max V i min


where Vimax and Vimin are the maximum and minimum time T in for a new generation call in system i is deter-
speeds of the MSs in the system i, respectively. The mined with the cell size and the mean speed of the
speed of the MSs in the given cell is uniformly dis- MS. The pdf of the cell residence time T h, i for a
tributed in [V imin, V imax]. Its mean speed is (V imin + handoff call in system i is in Appendix A.
V imax)/2 (Mar et al., 2007). The mean cell residence

V max + V min
, for t = 0
2π r i
2 2
4r i tV min tV max 2r i (18)
fTh(t) = 1– – 1– , for 0 ≤ t ≤ .
(V max – V min) π t 2 2r i 2r i V max
2
4r i tV min 2r i 2r i
1– , for ≤t≤
(V max – V min) π t 2 2r i V max V min
592 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

Handoff call to
adjacent cell

Compute
xj, Gj, N throughput
Set simulation Set simulation
interal T, vehicle interal T, vehicle Call
I(t) < Cx(t) ? RT
speed anc cell speed anc cell completion
size size
Call drop

Compute new call


blocking and
handoff failure
probabilities

Fig. 3 Simulation flow chart for throughput


The mean cell residence time T hi for a handoff call in total number of completion calls. The time subset in
system i is determined with the cell size and mean which blocking occurs is given by
speed of the MS. The simulation flow chart for chan-
nel throughput of Mobile WiMAX is shown in Fig. B = {t : I(t) < C B(t)}, (21)
3, where G is system capacity, T is total simulation
time, and x j and G j are the slot holding time and as- where C B(t) is the number of required slots for new
signed slots, respectively. The handoff rate is pro- generation calls at time t. The new call blocking prob-
portional to the reciprocal of mean cell residence time. ability in the given cell is obtained by
Therefore, the handoff rate at cell size x and mean
MS speed y for new generation call is determined by NB
PB = , (22)
Ng
n
T
R x,n y = Rrn nr , (19)
T x, y where the total number of blocked calls under the time
subset B is N B. The total number of new generation

where T x,n y is the mean cell residence time for new calls is defined as Ng. The time subset in which handoff
generation call under the conditions of cell size x and failure occurs in system i is written as follows:

mean MS speed y. Rrn and T rn are the reference handoff
rate and reference mean cell residence time, respectively, H = {t : I(t) < C H(t)}, (23)
for new generation call under the conditions of mean
MS speed 50 km/h and maximum cell size 1.52 km. where C H (t) is the number of required slots for
The handoff rate at cell size x and mean MS speed y handoff calls at time t. The handoff failure probabil-
for handoff call is given by ity in a given cell is written as

h NH
Tr PH = , (24)
R x,h y = Rrn h
, (20) Nh
T x, y
where the total number of handoff failure calls for

where T x,h y is the mean cell residence time of handoff time subsets H is defined as N H. The total number of

call at cell size x and mean MS speed y. T rh represents handoff calls is defined as N h. The channel through-
the reference mean cell residence time for handoff call put of Mobile WiMAX is determined with
under the conditions of mean MS speed 50 km/h and N
maximum cell size 1.52 km. For each new generation Σ
j=1
x jG j
call and handoff call, which can be served in the given RT = . (25)
T
cell if I(t) ≥ C x(t), where I(t) is the number of avail-
able slots at time t, Cx(t) is the number of the required where xj and Gj are the call holding time and assigned
slots for new generation calls or handoff call at time t. slots, respectively, for each new generation call and
The new generation calls or handoff calls will be dropped handoff call. In the preset simulation time T, the chan-
if I(t) < Cx(t). As soon as the call is generated in Mo- nel throughput R T can be calculated.
bile WiMAX (system i) successfully, the call may be The channel throughput of FUSC type downlink
completed or handed off to an adjacent cell. N is the Mobile WiMAX system utilizing three antenna modes,
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 593

Table 5 Channel throughput of multimedia call services in five environments for 50 Km/h, 100 Km/h and
120 Km/h MS speeds
Antenna modes case 3 (Fig. 2(a)) case 4 (Fig. 2(b)) case 3* (Fig. 2(c))
Rural Rural Rural Rural Rural Rural Rural Rural Rural
Environments Urban Suburban Urban Suburban Urban Suburban
A B C A B C A B C
(cell size (km)) (0.6746) (0.7522) (0.5539) (0.6176) (0.4548) (0.5071)
(1.09) (1.38) (1.52) (0.948) (1.19) (1.39) (0.823) (1.008) (1.094)
For
50 km/h 954.18 1052.69 1368.9 1818.12 1912.14 819.07 900.41 1300.74 1622.2 1864.4 1389.1 1533.1 2288.28 2832.61 2942.79
Throughput (kb/s)

speed
For
100 km/h 568.96 607.89 809.11 975.48 1079.2 503.15 543.68 718.8 879.9 980.7 918.3 957.6 1360.6 1521.44 1635.1
speed
For
120 km/h 513.18 550.21 710.44 864.66 923.45 458.14 496.66 631.1 725.8 865.26 886.1 895.15 986.1 1340.2 1423.05
speed

2500 2500

2000 2000
Throughput (kbps)

Throughput (kbps)

Urban
1500 Case 3 1500 Suburban
Case 4 Rural type A
1000 Case 3* 1000 Rural type B
Rural type C
500 500

0 0
0.18 0.19 0.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.18 0.19 0.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27
New call origination rate (calls/sec) New call origination rate (calls/sec)

(a) (b)

Fig. 4. Channel throughput of multimedia call services with varied new call origination rates (a) for three cases in urban environment and
at 50 Km/h MS speed; (b) for five environments in Fig. 2(a) and at 50 Km/h MS speed channel throughput in a given cell

in five environments, at three vehicle speeds (50 km/ antennas in all environments. Fig. 2(b) using 1 × 2
h, 100 km/h and120 km/h) is shown in Table 5, where SIMO antenna mode has the worst throughput due to
the call origination rate is set at 0.18 calls/sec; both diversity gain provided only in the receiver in all
the new call blocking probability and handoff failure environments. Under the same antenna mode, rural
probability are less than 0.05. The cell residence time type C has the better channel throughput than other
and handoff rate for new generation and handoff calls environments because it has the maximum cell size.
at different MS speeds are simulated and computed Nevertheless, the channel throughput of Mobile
using Eqs. (17), (18), (19) and (20), respectively. The WiMAX for a vehicle speed 50 km/h is less than twice
minimum handoff rates for new generation call are that of vehicle speed 100 km/h, though the call hold-
used as the reference handoff rate for the new gen- ing time doubles as the MS speed doubles because an
eration calls R nr. R nr is set as 0.1 to satisfy the condi- increase in handoff rate results in the increase of chan-
tions of both the new call blocking probability and nel throughput in a given cell. The channel through-
handoff failure probability less than 0.05. The chan- put of multimedia call services in five environments
nel throughput for three cases with varied new call for 120 Km/h MS speed is simulated to observe the
origination rates in urban environment, at 50 Km/h limitations of mobile WiMAX for deployment in
speed of MS is shown in Fig. 4(a) and the throughput Taiwan. Fig. 4(a) shows that the channel throughput
for five environments with varied new call origina- of the Mobile WiMAX increases with the call origi-
tion rates in case 3 and at 50 Km/h speed of MS is nation rate when the vehicle is driving in an urban
shown in Fig. 4(b). From Table 5, and from Figs. 4 area at 50 km/hr. Fig. 4(b) compares the channel
(a) and (b), it is noted that Fig. 2(c) using 2 × 2 MIMO throughput of Mobile WiMAX using 2 × 2 MIMO
SM antenna mode has the best channel throughput STC antenna mode in five different areas at 50 km/hr
due to different downlink data streams sent by each vehicle speed. It is noted that within these five areas,
transmission antenna; Fig. 2(a) using 2 × 2 MIMO the channel throughput of rural C is the best, followed
STC antenna mode has a medium throughput due to by rural B, rural A, and suburban, and that in urban
diversity gain provided in both transmitter and receiver area is the worst.
594 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

VI. CONCLUSIONS f the frequency in MHz


Fu Service reliability
In this paper, the cell size in a zone near Taipei g the all data-derived constants for each terrain
city for the BS of 2.5 GHz Mobile WiMAX system is category
planned. After the comparison of the propagation path G the system capacity
losses evaluated from the COST-231 Hata model and Gj the assigned slots
the SUI model, the COST-231 Hata model is proposed hb the BS antenna height above ground in meters
to estimate the cell size in urban and suburban areas, hr the MS antenna height above ground in meters
and the SUI Erceg model is proposed to estimate the H the time subset in which handoff failure oc-
cell size in rural A, rural B and rural C areas. The curs in system i
coverage probabilities for urban, suburban and rural I( . ) the number of available slots
environments are 70%, 76% and 82%, respectively, k the all data-derived constants for each terrain
where the service reliability is set as 90%. The link category
budget results validate that the cell size of the Mo- K the mean received signal strength in dB at cell
bile WiMAX using a 2 × 2 MIMO SM antenna mode boundary
is smaller than those using both 2 × 2 MIMO STC l the location in a cell
and 1 × 2 SIMO antenna modes. The cell size is de- L the radius of the location in a cell
termined with the link level performance of different N the total number of completion calls
radio setups including antenna mode, modulation and NB the total number of blocked calls under the
coding scheme, sub-channel type, fading margin, op- time subset B
eration environments etc. Based on the estimated cell Ng the total number of new generation calls
size, the traffic performance including the new call Nh the total number of handoff calls
blocking probability, handoff failure probability and NH the total number of handoff failure calls for
channel throughput are simulated and evaluated. The time subsets H
channel throughput simulation results show that us- p( . ) the probability density function
ing a 2 × 2 MIMO SM antenna mode can significantly PL basic pass loss
improve the channel throughput of the Mobile WiMAX. PB the new call blocking probability in the given
As soon as the vehicle speed increases from 50 km/h cell
to 100 km/h, the channel throughput of Mobile WiMAX P Wo( . ) the probability density function that the re-
reduces more than half because the call holding time ceived signal strength in dBm exceeds the
in a given cell decreases and the handoff rate increases threshold of the received signal strength in
upon increasing MS speed. dBm
RT the channel throughput in the preset simula-
ACKNOWLEDGEMENT tion time T
R rn the minimum handoff to adjacent cell ratio
This work was supported by ITRI and Far (handoff rate) for new generation call
EasTone Telecommunications Co., Ltd., Taiwan, R. s the shadow fading component
O.C. The authors would like to thank the editors and T the total simulation time
reviewers for their valuable inputs. Tu the unencumbered call duration
T h, i the cell residence time for a handoff call in
NOMENCLATURE system i
T n, i the cell residence time of for a new call
B the time subset in which blocking occurs T Hn, i the minimum of the unencumbered call dura-
cm the parameter defined as 0 dB for suburban tion in system i
or rural environments and 3 dB for urban en- T Hh, i the slot holding time for a handoff call in sys-
vironments tem i

C x( . ) the number of the required slots for new gen- T ih the mean cell residence time for a handoff call
eration calls or handoff calls in system i

CB( . ) the number of the required slots for new gen- T in the mean cell residence time for a new gen-
eration calls eration call in system i

CH( . ) the number of required slots for handoff calls T rh the reference mean cell residence time for
d the distance between receiver and transmit- handoff call under the conditions of mean MS
ter speed 50 km/h and maximum cell size 1.52
e the all data-derived constants for each terrain km

category T rn the reference mean cell residence time for new
erf ( . ) the error function generation call under the conditions of mean
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 595

MS speed 50 km/h and maximum cell size on Selected Areas in Communications, Vol. 17,
1.52 km No. 7, pp. 1205-1211.

T hx, y the mean cell residence time of handoff call Harri Holma and Antti Toskala, 2000, WCDMA for
at cell size x and mean MS speed y UMTS – Radio Access for Third Generaion Mo-

T nx, y the mean cell residence time for new genera- bile Communications, ISBN 0-471-72051-8, John
tion call under the conditions of cell size x Wiley & Sons, NY, USA, pp.157~159.
and mean MS speed y Hong, D., and Rappaport, T. S., 1986, “Traffic Sys-
Vi the speed of the MSs in system i tem and Performance Analysis for Cellular Mo-
V i max the maximum speed of the MSs in the system bile Radio Telephone Systems with Prioritized
i and Non-prioritized Handoff Procedures,” IEEE
V i min the minimum speed of the MSs in the system Transactions on Vehicular Technology, Vol. VT-
i 35, No. 3, pp. 77-92.
W the received signal strength in dBm IEEE Computer Society and IEEE Microwave Theory
Wo the threshold of the received signal strength and Techniques Society, 2006, IEEE Standard for
of the receiver in dBm Local and Metropolitan Area Networks Part 16:

W the mean of the received signal strength in Air Interface for Fixed and Mobile Broadband
dBm Wireless Access Systems Amendment 2: Physi-
x the zero-mean Gaussian variable of unity cal and Medium Access Control Layers for Com-
standard deviation N[0, 1] bined Fixed and Mobile Operation in Licensed
xj the slot holding time Bands and Corrigendum 1, 3 Park Avenue, NY,
y the zero-mean Gaussian variables of unit stan- USA, pp. 0_1-822.
dard deviation N[0, 1] Jakes, William C., 1993, Microwave Mobile Com-
z the zero-mean Gaussian variables of unit stan- munications, ISBN 978-0-7803-1069-8, Wiley-
dard deviation N[0, 1] IEEE Press, NY, USA, pp. 125-128.
Zi the distance between the point where a new Javornik, T., Kandus, G., Hrovat, A., and Ozimek, I.,
call is originated and the point on the cell 2006, Comparison of WiMAX Coverage at 450
boundary where an MS exits from cell MHz and 3.5 GHz, Department of Communica-
α the parameter in the equation of mean value tion systems, Jozef Stefan Institute, Jamova 39
of signal strength Ljubljana, Slovenia.
γ the parameter of a Gaussian random variable Liu, H., and Li, G., 2005, OFDM-Based Broadband
over the population of macrocells within each Wireless Networks Design and Optimization, John
terrain category Wiley & Sons, NY, USA, pp. 219-221.
λ the wavelength Mar, J., Chen, S. E., and Lin, Y. R., 2007, “The Ef-
µ the mean call unencumbered rate or call fect of the MS Speed on the Traffic Performance
completion rate of an Integrated Mobile WiMAX and DSRC Mul-
µσ the mean of the standard deviation of the timedia Networks on the Highway,” International
shadow fading component Conference on Communications and Mobile Com-
σ the standard deviation of the shadow fading puting Proceedings of the 2007 international con-
component ference on Wireless communications and mobile
σw the standard deviation of the received signal computing (IWCMC2007), pp. 44-48
strength in dBm Mar, J., and Huang, J. P., 2000, “Traffic Performance
σγ the standard deviation of γ Analysis of the Integrated Dual Band Cellular Radio
σσ the standard deviation of the standard devia- Networks,” IEE Proceedings Communications, Vol.
tion of the shadow fading component 147, No. 3, pp. 180-186.
Nelson, R., Probability, 1995, Stochastic Process and
REFERENCES Queueing Theory, Spriber-Verlay, NY, USA.
O’Reilly, P., 1999, Simulation with Visual SLAM and
Abhayawardhana, V. S., Wassell, I. J., Crosby, D., AweSim, Wiley, NY, USA.
Sellars, M. P., and Brown, M. G., 2005, “Com- Parsons, J. D., 1992, The Mobile Radio Propagation
parison of Empirical Propagation Path Loss Models Channel, ISBN 0-471-96415-8, John Wiley &
for Fixed Wireless Access Systems,” 61st IEEE Sons, NY, USA, pp. 156-159.
Vehicular Technology Conference, Stockholm, Rappaport, Theodore S., 2002, Wireless Communi-
Sweden, Vol. 1, pp. 73~77. cations – Principle and Practice, 2nd ed., Prentice
Erceg, V., and Greenstein, L. J., 1999, “An Empiri- Hall PTR, NJ, USA, pp. 138-141.
cally Based Path Loss Model for Wireless Chan- Varberg, D., and Purcell, E. J., 1997, Calculus,
nels in Suburban Environments,” IEEE Journal International 7th ed., Prentice-Hall Inc., NY,
596 Journal of the Chinese Institute of Engineers, Vol. 32, No. 5 (2009)

USA, pp. 443. in the approximating circle cell of radius r i. The pdf
WiMAX Forum, 2007, A Comparative Analysis of of the distance between the point where a new call is
Mobile WiMAX Deployment Alternatives in the originated and the point on the cell boundary where
Access Network, pp. 16-19, Available: http:// the MS leaves the cell was derived in (Hong and
www.wimaxforum.org/technology Rappaport, 1986).
WiMAX Forum, 2006, Mobile WiMAX – Part I: A Tech-
nical Overview and Performance Evaluation, pp. 30-
r i2 – z
2 2
33, Available: http://www.wimaxforum.org/ , for 0 ≤ z ≤ 2r i ,
fZ i(z) = π r i2 2
(A2)
technology. i = A or B ,
0, elsewhere .
Manuscript Received: Oct. 30, 2007
Revision Received: Oct. 23, 2008 Then the cell residence time for a new call is ex-
and Accepted: Nov. 23, 2008 pressed by T n, i = Z i/V i with pdf

APPENDIX A DERIVATIONS OF fTn, i(t)


Eqs. (17) AND (18)

The pdfs of the cell residence time for a new = w fz i(tw) fV i(w)dw
–∞
call and a handoff call were derived in (Hong and
Rappaport, 1986) where the minimum MS speed is
V i max 2
zero. Here we derive the pdf of the cell residence 2 w r i2 – tw dw ,
time for non-zero minimum MS speed, i.e., V i is uni- (V i max – V i min)π r i2 V i min 2
formly distributed on the interval [Vi min, Vi max], where 2r i
for 0 ≤ t ≤ ,
V imin is positive and less then V i max. V i max
=
i 2r /t 2
2 w r i2 – tw dw ,
1 , for V i min ≤ vi ≤ V i max , (V i max – V i min)π r i V i min
2 2
V i max – V i min
fV i(vi) = 2r i 2r i
i = A or B , for ≤t≤ , i = A or B .
0, elsewhere . V i max V i min
(A 1)
(A 3 )
It’s assumed that the angle between the direction of
the MS and the direction from the MS to the center Using L’Hô pital’s Rule (Varberg and Purcell, 1997),
of the cell is uniformly distributed in interval [0, π ], Eq. (A 3 ) becomes
and its direction remains constant during its travels

V i max + V i min
πri for t = 0 ,

2 3 2 3
8r i tV i min tV i max 2r i
1– – 1– , for 0 ≤ t ≤ ,
fTn, i(t) = 3(V i max – V i min)π t 2 2r i 2r i V i max
(A 4 )
2 3
8r i tV i min 2r i 2r i
1– , for ≤t≤ ,
3(V i max – V i min)π t 2 2r i V i max V i min
i = A or B .

For the handoff call, the angle between the direction leaves the cell was given in (Hong and Rappaport, 1986).
of the MS and the direction from the MS to the center
of the cell is assumed to be uniformly distributed in
1 1 , for 0 ≤ z ≤ 2r i ,
π
interval [– π , π ]. The pdf of the distance between the r i2 – z
2
2 2 fZ i(z) = 2 (A5)
point on the cell boundary where a handoff call is origi- i = A or B ,
nated and the point on the cell boundary where MS 0, elsewhere .
J. Mar et al.: Cell Planning and Channel Throughput of Mobile WiMAX at 2.5 GHz 597

Then the cell residence time T h, i for handoff call is expressed by Th, i = Zi /Vi. The pdf of Th, i is derived as

V i max + V i min
, for t = 0 ,
2π r i
2 2
4r i tV i min tV i max 2r i
1– – 1– , for 0 ≤ t ≤ ,
fTh, i(t) = (V i max – V i min)π t 2 2r i 2r i V i max (A 6)
2
4r i tV i min 2r i 2r i
1– , for ≤t≤ ,
(V i max – V i min)π t 2 2r i V i max V i min
i = A or B .

If V i min = 0 is substituted into Eqs. (A4) and (A 6) the (Hong and Rappaport, 1986) are obtained.
identical formulas of f T n, i(t) and f T h, i(t) as shown in

You might also like