Performance of HSDPA and HSUPA at 900/2000 MHz bands

1

Performance of HSDPA and HSUPA at 900/2000

MHz bands

João Pedro Roque1, Sérgio Pires

2 and António Rodrigues

1

1Instituto de Telecomunicações / Instituto Superior Técnico

Technical University of Lisbon, Lisbon, Portugal 2Celfinet, Portugal

Abstract— The main purpose of this paper is to study the High

Speed Downlink Packet Access (HSDPA) and High Speed Uplink

Packet Access (HSUPA) performance at 900/2000 MHz bands,

considering a multiple users and services scenario. The intention

is to evaluate the traffic management between the two carriers,

the coverage and capacity aspects. A simulator was developed to

study the multiple users’ scenario, enabling the analysis of

network performance by varying several parameters. In order to

better evaluate the system HSDPA and HSUPA 900/2000 MHz,

two different strategies were applied in the network, the “Carrier

2000 Loading” and the “Priority Service” strategy.

Keywords - UMTS, HSDPA 900/2000 MHz, HSUPA 900/2000

MHz, Traffic Management Strategies, Capacity, Coverage.

I. INTRODUCTION

UMTS carriers are currently used by data services based on the HSDPA and HSUPA technologies, which can now deliver peak data rates of up to 7.2 Mbps and 1.45 Mbps, respectively. The coverage of these services is unfortunately limited in the standard UMTS 2000 MHz band. This, results in a significant reduction of the practical data rates that can be delivered, especially inside large buildings, as well as in rural areas. One cost-effective way to address this issue is to deploy the mobile broadband services in the Global System for Mobile Communications (GSM) 900 MHz band, where the propagation characteristics are much more favourable when compared to the regular 2000 MHz band. This solution is referred to UMTS 900.

Furthermore, the UMTS deployed at 2000 MHz frequencies, where the signal attenuation is higher than at 900 MHz, requires fairly high site density making a challenge to match existing GSM 900 coverage with the same sites. Thus, UMTS can take benefit of the better signal propagation at 900 MHz, improving the indoor coverage and the cell sizes.

Hence, the main purpose of this paper is to study HSDPA and HSUPA performance at 900/2000 MHz bands, considering a multiple users and services scenario. The intention is to evaluate the traffic management between the two carriers, the coverage and capacity aspects, such as average network throughput and satisfaction rate, among others. A comparison of HSDPA performance for 5, 10 and 15 HS-PDSCH codes is addressed, as well as the different capacity impacts of the increase of the number of codes. The objectives were accomplished through the development and implementation of

a simulator that enables the analysis of HSDPA and HSUPA multiple users’ model at 900/2000 MHz, being capable to produce results according to several parameters.

The remainder of this paper is outline as follow. Some basic concepts necessary to understand the UMTS, HSDPA, HSUPA and UMTS 900 technologies are explained in section II. In section III, it is presented the multiple users’ model. In section IV, the results analysis is described. Finally, in section V, the conclusions are draw and the future research purposed.

II. BASIC CONCEPTS

A. UMTS

Since the beginning, UMTS network has been designed to

support any type of services, where each service does not

require particular network optimisation, whereas the 2nd

generation systems were designed for efficient delivery of

voice services. The UMTS network must be able to deliver

high and reasonably constant bit rate, to avoid high delay

connections. These bit rate and delay requirements may be

achieved in a cost efficient way by utilising the Quality of

Service (QoS) differentiation features that are available in

UMTS. Thus, when the system load is getting higher, it

becomes more important to prioritise the different services

according to their requirements, through the QoS.

B. HSDPA

The HSDPA concept was designed with the purpose to

improve the downlink (DL) packet data throughput, being

deployed on the top of WCDMA network and launched in

March 2002. HSDPA is also capable to improve capacity and

spectral efficiency, sharing all network elements with Release

99.

The HSDPA performance depends significantly on the

network algorithms, deployment scenarios traffic, QoS and

mobile terminal (MT) receiver performance and capability.

There are 12 MT categories in HSDPA, being the achievable

maximum data rates ranging from 0.9 to 14.4 Mbps.

C. HSUPA

HSUPA specification work started with focus on evaluating

potential enhancements for the uplink (UL) dedicated

2

transport channels, after the successful finalization of the first

version of HSDPA. Hence the first version was launched in

December 2004 by the 3rd Generation Partnership Project

(3GPP), being known as HSUPA Release 6.

HSUPA emerged to improve capacity and data rates in the UL

direction, being possible to achieve 1 to 2 Mbps data rates

compared to Release 99’s 384 kbps.

This technology uses most of the basic features of Release 99

in order to work, such as power control loop and Soft

Handover (SHO) which are essential for HSUPA operation.

The only change is a new way of delivering user data from the

user equipment (UE) to the Node B.

Similar to HSDPA, performance in HSUPA depends on

parameters such as network algorithms, deployment scenario,

MT transmitter capability, Node B receiver performance and

capability and type of traffic. Thus, there are 6 MT categories

for HSUPA, being the achievable maximum data rates ranging

from 69 kbps to 4.059 Mbps

D. UMTS 900

The main feature of UMTS 900 is to provide a better indoor

coverage in existing UMTS deployment areas and enabling

larger cell sizes in new UMTS areas, being possible to use the

same services and the same peak data rates as used by UMTS

2000.

Hence, deploying UMTS 900 with HSPA (High Speed Packet

Access) by reusing the existing GSM sites allows mobile

operators to offer UMTS services, such as high data rate

multimedia services, to the benefits of the consumers.

When there is a hot spot, e.g. tourist places, train stations, etc,

where more capacity is needed, higher frequency bands, such

as 2000 MHz band, can be used to offer additional capacity, as

shown in Figure 1.

Figure 1. UMTS coverage at the 900MHz and 2000 MHz,

extracted from [4].

Actually, the cell area with UMTS 2000 is from 2.5 to 3.0

km2, while in UMTS 900 the cell area can be 2.5 times larger,

7 to 8 km2. On the other hand, UMTS 900 can reduce the

required number of base station sites by 60 %, while

maintaining the same coverage [1], as illustrated in Figure 2.

The UMTS 900 voice and data have higher coverage, when

compared with UMTS deployed at 2000 MHz frequencies.

Hence, the deployment of less base station sites directly

implies lower cost for the network.

Figure 2. Suburban cell size with 95 % indoor coverage,

extracted from [2].

Concerning the carrier separation, when deploying macro

cellular UMTS 900 in urban area and rural area in co-

existence with another UMTS 900 network, the carrier

separation between two UMTS networks should be 5 MHz or

more, similar to the UMTS deployment in 2000 MHz band.

Concerning the traffic management between UMTS 900 and

UMTS 2000, it is believed that all, or at least, most of UMTS

900 handsets will have UMTS 900/2000 dual-band capability.

Taking into account that UMTS 900 cell range is larger than

UMTS 2000, in a dual-band UMTS 900/2000 network UMTS

traffic should be sent to UMTS 2000 layer (a). When possible,

leaving UMTS 900 layer to handle traffic in the area where

there is no UMTS 2000 coverage (b), as shown in Figure 3.

Figure 3. Traffic management between UMTS 900 and UMTS

2000, extracted from [4].

As a conclusion, the most significant benefit of deploying

UMTS in 900 MHz frequency band comes from the fact that,

compared to 2000 MHz band, radio wave propagation pathloss

at 900 MHz is much smaller. So, offering the same service

(data rates) and same coverage, the required number of sites in

900 MHz band is reduced by 60% compared to that in 2000

MHz band, [4]. This will bring economic benefit on UMTS

operator’s investments and makes it possible to propagate

benefits to the end-users in terms of wider coverage and

possibly lower level of usage costs. UMTS 900 will be

deployed by reusing the GSM sites within the existing service

area. Deploying UMTS 900 with HSPA in rural area by

reusing the existing GSM sites is a cost-effective solution for

mobile operators to offer UMTS services, such as high data

rate multimedia services, whereas deploying the UMTS at 900

MHz band in urban areas can improve indoor coverage.

3

III. MODELS AND SIMULATOR DESCRIPTION

1) HSDPA and HSUPA models

In this section, a description of the HSDPA and HSUPA 900/2000 MHz model for the analysis of traffic management between Node Bs of 900 MHz and 2000MHz is presented. The objective of this model is to know the network capacity, average satisfaction rate and the instantaneous throughput available, according to the topology introduced in the user interface, which allow to modify several parameters, such as, Node B transmission power, number of HS-PDSCH codes for HSDPA simulator, Node Bs and MT antenna gains, type of environment, among others.

Regarding this analysis, two different strategies were developed considering the same network deployment, being necessary to calculate the maximum cell radius for both cells, 900 MHz and 2000 MHz. The maximum cell radius is calculated for the maximum distance that allows the user to be served with the desired throughput, so it was calculated for the service with higher throughput.

The pathloss is calculated using the COST-231 Walfisch-Ikegami propagation model:

[ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ]dBdBirdBmrdBmdBtmdBttdBdBp MGPEIRPLLLL −+−=++= 0 (1)

where, L0 is the free space loss; Ltt, the rooftop-to-street diffraction loss; Ltm, the approximation for the multi-screen diffraction loss; EIRP, the Equivalent Isotropic Radiated Power; the Pr, available receiving power at the antenna; Gr, the receiving antenna gain; M, the total margin.

Thus, through the manipulation of (1) and the Ltt and L0

expressions from the propagation model, the maximum cell

radius can be calculated by the expression:

[ ]

[ ] [ ] [ ] [ ] [ ] [ ] [ ]

d

dBdBtmdBttdBdbiRdbMRdBm

K

LLLMGPEIRP

KmR+

−−−−+−

=20

'0

'

10

(2)

where, [ ])log('

Kmdtttt dKLL −= ; dK is the dependence of the

multiscreen diffraction loss versus distance; d is the distance

between the user and the Node B; [ ])log(200

'

0 KmdLL −= and R

is the maximum cell radius.

Regarding frequency, one can conclude that for 900 MHz, the

cell radius for the maximum service throughput considered in

HSDPA, 2.048 Mbps, is approximately 0.6 km, whereas the

cell radius for 2000 MHz is 0.3 km. This is due to fact that

with the increase of the frequency, the path loss increases,

leading to a cell radius decreases.

Regarding the traffic management optimization between the

two types of cells, it is considered one of the two strategies

available in the simulator, “Carrier 2000 Loading” or “Priority

Service” strategy. Furthermore, the operators can use several

strategies depending of the usersP importance, services’

priority, or different strategies can simply be applied for day

and for night. Following this approach, it was considered the

“Carrier 2000 Loading” to represent the most common

strategy, independent of if it is a dense area, if it is day or

night. This strategy takes into account the services’

penetration percentages and the priority list. The “Priority

Service” strategy also takes into account these same features,

but uses it in a different way, emphasizing the most priority

service. This second strategy is most useful when the network

is overflow and the operators opt to guarantee that the most

priority users are served.

In the “Carrier 2000 Loading” algorithm all users are

connected to the 2000 MHz carrier at the beginning. After, the

system capacity is analysed at each Node B, by summing the

throughput of all users connected to it. If the sum is higher

than the throughput threshold, the users are moved to the 900

MHz Node Bs, one by one, according to their distance and the

QoS priority list, since the furthest and less priority users are

the first to be analysed. Then, the capacity analysis is once

again made and if the system is over limit, the reduction

strategy at the 900 MHz Node Bs level is applied.

In the “Priority Service” strategy algorithm, initially, the users which belong to the most priority service are connected to the 900 carrier, whereas the others users are connected to the 2000 MHz Node Bs. Then, the system capacity is analysed at the Node Bs, by summing the throughput of all users connected to them, being the 2000 MHz cells the first to be analysed. Thus, if the total throughput is higher than the limit of each Node B, the users are moved to the 900 MHz cell, one by one, considering the distance and the QoS priority list. Finally, it is applied the reduction strategy, if the 900 MHz Node Bs are over their limit.

2) Network Deployment

The simulator has the capacity to load different scenarios,

according to the previously set parameters (e.g. physical

conditions, antenna configurations and type of topology),

giving the application a wide range of deployments and

outputs, due to the combination of the these parameters in

several ways.

Thus, it is possible to analyze the influence of a specific

parameter in the network, in order to evaluate the HSDPA and

HSUPA 900/2000 MHz performance. The application places

the 900 MHz Node Bs co-sited with the 2000 MHz, being the

number and the localisation of each one read from the “txt”

file set in the user interface, which takes into account the

maximum cells radius. The co-sited Node Bs are alternate

with the 2000 MHz Node Bs, along the network dimensions,

since the 900 MHz cells are bigger than the 2000.

3) User’s Generation

The main objective of the user’s generation module is to

generate and distribute a number of users, set at the user

interface, according to the services percentages defined in

Table 1. The services considered in this simulator do not

4

include voice; they are only data services, since the HSDPA

and HSUPA are mainly used for this type of services.

Table 1. Profile One Characterisation.

Penetration

Percentage [%] Services

Profile 1

Quality of Service (QoS)

priority

Web 46.4 1

Streaming 6.2 2

E-mail 1.0 3

FTP 1.0 4

Chat 3.1 5

P2P 42.3 6

Firstly, the number of users that belong to one of the services

considered in Table 1, are obtained using the Matlab random

function, being generated a number between 0 and 1.

This random number lays in one of the intervals that takes into

account the services’ penetration percentages, and according

to this interval the MS is set to a specific service. It is also

taken into account the users’ priority, according the QoS

priority list presented. These values can also be consulted in

Table 1.

Furthermore, this module has the capacity to distribute the

users’ generated, using a uniform distribution, for the network.

The distribution takes to account the network dimensions,

since the generated numbers are distributed by the network

width and height dimensions, for the horizontal and vertical

positions, respectively.

4) HSDPA and HSUPA 900/2000 MHz Implementation

The HSDPA and HSUPA are two different modules, which

were implemented to enable the analysis of the impact of the

900 MHz band in the network. However, this two applications

run under the same Network Deployment and make use of the

same Users’ Generation.

The HSDPA and HSUPA were built to analyze the network

capacity and coverage, through a snapshot approach,

calculating instantaneous network results as number of user

per Node B and traffic. Since they have a similar function,

these modules will be analysed together.

Concerning the main objectives of the simulator, the need to

calculate the signal to noise ratio (SNR) is inherent and

mandatory for the both applications. Using the throughput in

function of SINR graphic, the requested throughput is mapped

into SINR and Ec/N0 for HSDPA and HSUPA, respectively.

Hence, it is possible to calculate the minimum received power

that allows the user to be served with the requested

throughput, calculation.

min[dBm] [dBm] [dB] [dB]Rx P

P N G SNR= − + (3)

where, N is the Total Noise Power; Gp, the processing gain;

SNR, the Signal to Noise Ratio.

Firstly, the pathloss is calculated using the expression shown

in (1). Besides the considered path loss, the total pathloss

depends on some additional margins, which are set in the

menu interface with the objective to simulate different channel

conditions. These margins are related to the slow and fast

fading margins, being described by a Gaussian and Rayleigh

distribution, respectively, to the indoor penetration losses and

to the soft handover gain for HSUPA, being the impact on the

overall results quite significant.

After the total pathloss, several parameters related to the

Equivalent Isotropic Radiated Power (EIRP) are requested,

being set by the user. The EIRP requested parameters are: BS

and MT transmission power; BS and MT antenna gain; User

losses; Cable losses; Diversity gain, only for HSUPA; System

signalling and control power.

Another, important physical aspect is the noise power, which

is calculated taking into account the following parameters also

obtained from the menu interface: Noise Factor and the

Interference Margin. The interference margin is calculated

based on the total number of users of the Node B coverage

area. In this paper it is considered the higher number of users

connected to the Node Bs, thus the interference margin taking

into account for HSDPA and HSUPA is the maximum value,

i.e., 6 dB.

The number of HS-PDSCH codes is an important parameter

for the evaluation of the network HSDPA performance, since

it is responsible for the throughput increase of the end of user.

Finally, the application calculates the receiver sensitivity with

the objective to obtain SNR values of each user in the

network.

[ ] [ ] [ ] [ ]

[ ] [ ] [ ] [ ] [ ] [ ]dBdBdB

dBdBm

pdBmcudBirpdBm

pdBmRxdB

GNLGLEIRP

GNPSNR

+−−+−=

=+−=

/

(4)

After calculating the SNR values, the throughput is associated

with each user, according to the user distance between all

Node Bs to all users, i.e, the maximum throughput that each

Node B can offer to each user, using the expressions of

interpolation curves from [3]. Then, the user is connected to

the closest Node B, which has the minimum pathloss, and is

associated with the throughput of that connection.

Furthermore, concerning the throughput that can be offered to

the user, it is necessary to take into account 3 different

situations:

i) When the throughput associated to the distance is higher

than the service’s throughput, the user is served with the

requested throughput;

ii) On the other hand, if the throughput distance is higher than

the minimum service and lower than the maximum service

throughput, the user is served with the throughput distance;

iii) The other situation is when the throughput distance is

lower than the minimum service throughput, being the user

without coverage and is counted “outage”.

5

Additionally, the analysis of the system’s capacity is carried

out at the Node Bs, by summing the throughput of all users

connected to the Node B. So, if the sum is lower than the

maximum allowed throughput for Node B, all users are served

without reduction, whereas if the sum is higher than the

throughput threshold in Table 2, the “Quality of Service

Reduction” reduction strategy is applied. In this reduction

strategy all the users’ throughput of the same service is

reduced by 10%, according to a list containing the services’

priorities.

Table 2. HSDPA and HSUPA maximum application

throughput.

System Maximum Throughput

[Mbps]

HSDPA 5 HS-PDSCH codes

3.0

HSDPA 10 HS-PDSCH

codes 6.0

HSDPA 15 HS-PDSCH

codes 8.46

HSUPA 1.22

The approach used for HSDPA is also used for HSUPA, being

the SHO the only difference, because each user can be

connected to the two closest’ Node Bs. In this case, the user

throughput is the minimum throughput allowed by one of the

two available throughputs from each Node Bs. It is also

considered a limit for a user to be in SHO, 0.384 Mbps, since

a user in SHO allocated more resources than when the user is

connected to only one Node B.

The services considered in this simulator do not include voice;

they are only data services, since the HSDPA and HSUPA are

mainly used for this type of services. The Profile One

characterisation is shown in Table 3.

Table 3. Profile One Characterisation.

Penetration Percentage [%] Services

Profile 1

Quality of Service (QoS) priority

Web 46.4 1

Streaming 6.2 2

E-mail 1.0 3

FTP 1.0 4

Chat 3.1 5

P2P 42.3 6

IV. RESULTS ANALYSIS

As the main objective of this paper is to study HSDPA and

HSUPA performance at 900/2000 MHz bands, a set of

simulation scenarios was conceived in order to evaluate this

impact, regarding several parameters variation. Hence, it was

adopted a default scenario to simplify the results analysis, over

which was performed the parameters variation, comparisons

and further conclusions. The default scenario has the goal to

study the multiple users’ scenario, which considers the users

uniformly distributed along the network, performing different

services with different throughputs.

The main objective of the following analysis is to show how

much the use of the 900 MHz carrier could have improved the

previous releases, and to pick the systems with better

performance for DL and UL, which are the ones from the later

releases. For these systems, simulations were performed, for

DL and UL, varying the available configurations, with the

objective to evaluate the relative impact, in terms of coverage,

capacity and throughput.

The parameters considered for the default scenario are the 15

HS-PDSCH codes for HSDPA, 1000 users and the Profile

One. Then, it is analysed some parameters variation, such as

the 5 and 10 HS-PDSCH codes for HSDPA, 3000 users and

two others profiles characterization, as it is shown in Table 4.

Table 4. Alternative profiles characterisation.

Penetration

Percentage [%]

Penetration

Percentage [%] Services

Profile Two Profile Three

Quality of

Service (QoS) priority

Streaming 10 10 1

Web 40 40 2

Chat 10 20 3

FTP 10 10 4

P2P 10 5 5

E-mail 20 15 6

A. HSDPA 900/2000 MHz evaluation

All the results presented in this subsection were obtained using

the simulator and the HSDPA model introduced in Section III.

Hence, several simulations were effectuated for the two traffic

management strategies.

1) Default Scenario

Considering all the users served in all simulations performed

and their distance to the Node B they are connected to, it was

evaluated the instantaneous user throughput for the two

carriers. From Figure 4 one can notice that for a distance

further than 0.5 km, the user throughput starts to have an

irregular behaviour for both carriers. This occurrence can be

explained by the decreasing number of users that are served

when the user’s distance increases, since according to the

“QoS Reduction” strategy, the users are also evaluated from

their distance, i.e, the throughput reduction strategy starts to

be applied from the furthest user to the closest one. Hence, the

number of delayed and outage users are higher for distance

above 0.5 km.

Furthermore, the main differences of the user throughput for

the 900 MHz carrier and 2000 MHz carrier are due to the

traffic management strategy adopted in this case. For the 2000

MHz carrier, it is possible to notice three lines, where the

users are mainly concentrated. This fact happens, because the

users are firstly set to the 2000 MHz BSs, being moved to the

6

900 BSs only after the system capacity analysis. Thus, there

are few users reduced and the higher number are localised

around their requested throughput, which are 1.536, 1,024 and

0.384 Mbps. Concerning the 900 MHz carrier, Figure 4 left

side, it is illustrated the “QoS Reduction Strategy”

performance, since the users are mainly spread between the

0.1 km and 0.5 km and are served with a large variety of

throughputs, according to the number of reductions suffered

by them.

0 0.2 0.4 0.6 0.80

0.5

1

1.5

2

Distance [Km]

Instantaneous User Throughput [M

bps]

900 MHz Carrier

0 0.2 0.4 0.6 0.80

0.5

1

1.5

2

Distance [Km]

Instantaneous User Throughput [M

bps]

2000 MHz Carrier

Figure 4. HSDPA instantaneous user throughput for all users

depending with distance, according to users’ distribution one.

The instantaneous user throughput for all network was got

from the overlap of both carriers, Figure 5. It can also be

noticed the same behaviour of each carrier separately, which is

also justified by the fact that the average user distance is

around 0.3 km.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.90

0.5

1

1.5

2

Distance [Km]

Instantaneous User Throughput [M

bps] HSDPA Instantaneous Throughput for all users

Figure 5. HSDPA instantaneous users’ throughput of all

network depending with distance.

Regarding the offered and served traffic, there is a small

reduction of Web users which is compensated by a small

increase of 3.54% on the percentage of P2P users. The offered

and served percentages are similar for the other services

analysed. Due to the QoS differentiation introduced in the

reduction strategy used, the service with the higher priority,

Web, has the highest average instantaneous throughput per

user and the second highest satisfaction rate, almost 80%.

The average network throughput is 6.64 Mbps, while the

average instantaneous user throughput is around 0.81 Mbps

with a satisfaction rate of around 64.69%. Relatively to the

average instantaneous throughput per user, it is possible to

conclude that it decreases with distance, since the signal to

interference noise ratio (SINR) is the limiting factor, due to

the introduction of the interference margin in the multiple

users’ scenario. Thus, the SINR value for a user further away

from the Node B becomes lower, leading to a reduction of the

throughput given to each user. Further, there are differences

between the 900 MHz carrier and 2000 MHz carrier, due to

the traffic management strategy adopted in the default

scenario. This difference is in the user throughput, where in

the 2000 MHz carrier it is possible to notice that the users

spread over their requested throughput, since they are only

reduced in 900 MHz carrier.

2) Number of HS-PDSCH Codes

The influence of the number of HS-PDSCH codes was also

analysed, and can be observed that an increase from 10 to 15

codes improves the average network throughput around 2.14

Mbps, i.e. 32%, as well as the maximum throughput allowed

for a single Node B, since there are more codes available for

data transmission. When the variation is from 5 to 15 codes,

the increasing in average network throughput, satisfaction rate

and total served users is obviously higher, i.e. around 68%,

40% and 39% respectively. As conclusion, more codes

available lead to an increase of the total network throughput,

satisfaction rate and number of served users.

3) Number of Users

Relatively to the number of users’ impact in the network, it

can be seen that when introducing more users, from 1000 to

3000 users, the average network throughput presents an

increase of 13 %, as there are more users in the network to be

served. This increase is mostly due to the higher number of

web users served, which are the most priority and have a

higher requested throughput. Hence, the majority Node Bs are

full with the web users with their maximum throughput due to

the “Carrier 2000 Loading” strategy. On the other hand, as

there are more users in the network, and the resources are the

same, there is a reduction in the average satisfaction rate of

approximately 21 %. With more users in the network, the

average satisfaction rate and average ratio of served users

decrease, because the resources are the same to serve more

users. The average ratio of served users is lower, but the

number of effective served user when considering 3000 users

is higher. As expected, this difference is due to the fact that

the users suffer a higher reduction in the 900 MHz carrier,

when considered 3000 users, whereas the number of users in

2000 MHz carrier is approximately the same.

4) Alternative Profiles

When comparing the Profile One with Profile Two and Three,

Table 3, the alternative ones present a significant reduction on

the percentage of users performing P2P, which is one of the

most demanding services in terms of throughput. On the other

hand, there is an increase in the percentage of users

performing Chat, Email and FTP. In an overall perspective, it

can be said that both alternative profiles are more demanding,

in terms of users’ throughput, than the Profile One.

Evaluating these variations, it is possible to notice that there

are not variations on the average network throughput for the

alternative profiles, when comparing with the Profile One. For

the average satisfaction rate, there is a decrease when

changing from the Profile One to the alternative profiles. The

7

profile number one has a significant number of users

performing Web, which was move from first to second in the

priority list, being, reduced more users of this service than in

the Profile One, leading to a lower average satisfaction rate.

The percentage of users performing Chat, whose maximum

throughput considered, is 0.384 Mbps, increases. These users

can be more easily served by the network, since are applied

less reductions. This implies an overall increase of the average

satisfaction rate from the Profile Two to the Profile Three.

5) Strategies Comparison

Analysing the two different strategies, which were applied in

the network, the “Carrier 2000 Loading” and the “Priority

Service” strategy, it is possible to notice that the “Carrier 2000

Loading” strategy is the best one, as it allows a higher

throughput and satisfaction rate per service in general, except

for web service, which presents better results when it is used

the “Priority Service” strategy.

The “Carrier 2000 Loading” strategy improves 0.33 Mbps the

average network throughput, representing an increase of 5.2

%, and it allows an increasing of 2.68% in the satisfaction

rate, when compared with the “Priority Service” strategy.

On the other hand, the “Priority Service” strategy is the best

one when the operators want to give extremely importance to

one specific group of users. As illustrated in this paper, it is

possible to improve the Web average throughput and

satisfaction rate, in 2.4% and 2.3% respectively, by using the

“Priority Service” instead of using the “Carrier 2000 Loading”

strategy.

B. HSUPA 900/2000 MHz evaluation

All the results presented in this subsection were obtained using

the simulator and HSUPA model introduced in Section III.

Hence, it was evaluated the instantaneous user throughput of

all network and for the both carriers separately.

1) Default Scenario

From Figure 6, is possible to notice the expected differences,

which are due to the “Carrier 2000 Loading” strategy. The

user can only be served by the 2000 carrier for distances until

0.38 km, as is shown in Figure 6 left side, since the HSUPA

system has less available resources than the HSDPA. Thus, the

users requesting HSUPA services for distances furthest than

0.38 km are moved to the 900 MHz Node Bs due the “Carrier

2000 Loading” strategy. On the other hand, it is possible to

observe the same three lines present in HSDPA user

throughput, being here where the users are mainly concentrate,

since they are localised around their requested throughput

which are 0.512, 0.384 and 0.064 Mbps.

0 0.2 0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

Distance [Km]

Instantaneous User Throughput [M

bps] 900 MHz Carrier

0 0.2 0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

Distance [Km]

Instantaneous User Throughput [M

bps] 2000 MHz Carrier

Figure 6. HSUPA instantaneous user throughput for all users

depending with distance, according to users’ distribution one.

As it is shown in Figure 7, the average instantaneous user

throughput is approximately constant for distances until 0.5

km. After this distance the network behaviour tends to be

irregular, which is due to the fact that few users are served

with the requested throughput when they are placed beyond

0.5 km, since according to the “QoS Reduction” strategy, the

users are evaluated from their distance, i.e, the throughput

reduction strategy start to be applied from the furthest user to

the closest one. The use of SHO also helps to justify the

constant throughput results, as when the distance increases and

one would expect a reduction of the user throughput, the

probability of the user being in SHO increases, as the user is

more likely to be near the cell edge. For the default scenario,

the average network radius is 0.23 km.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.90

0.1

0.2

0.3

0.4

0.5

Distance [Km]

Instantaneous User Throughput [M

bps] HSDPA Instantaneous Throughput for all users

Figure 7. HSUPA instantaneous users’ throughput of all

network depending with distance.

Regarding the offered and served traffic, it is possible to

conclude that the reduction of 2.12% in the served users

performing Streaming, is mainly due to fact that these users

requested throughput is equal to its minimum throughput,

meaning that, when reductions are performed more than one

time, Streaming users are always delayed.

2) Number of Users

Relatively to the number of users’ impact in the network for HSUPA 900/2000 MHz, it can be seen that when introducing more users, from 1000 to 3000 users, the average network throughput presents an increase of 18.4%. With more users in the network, the average satisfaction rate and average ratio of served users decrease, because the resources are the same to serve more users. The average ratio of served users is lower, but the number of effective served user when considering 3000 users is higher.

8

3) Alternative Profiles

When introducing more demanding profiles, such as Profile

Two and Profile Three, the network is still capable of serving

the same users as the ones for the Profile One. This fact

explains the approximate same value for the percentage of

total served users, around 37%, i.e. 370 users. As for the

average satisfaction rate, there is a smooth increase when

changing from the Profile One to the alternative profiles. This

is due to the fact that the first one has a significant number of

users performing P2P, which is the first service to be reduced,

leading to a lower average satisfaction rate. The alternative

profiles present a low percentage of P2P, which it is now the

second service to be reduced. There is also an increase in

terms of percentage of Chat, whose maximum throughput is

0.384 Mbps, which can be more easily served by the network,

therefore, less reductions have to be performed. Concluding

implies an overall increase of the average satisfaction rate for

the alternative profiles.

For HSUPA, there are an increase of 0.5% and 0.6%, from

Profile One to Profile Two and Three, respectively.

4) Strategies Comparison

As observed in HSDPA, the “Carrier 2000 Loading” strategy

has a better performance in HSUPA, since allows uniform

results for all services. In the service by service analysis,

Streaming has an average satisfaction rate of 100%, even

though Web is the service with the highest QoS priority. This

is due to the fact that, for HSUPA, the maximum and

minimum Streaming throughputs are equal, meaning that,

when the Streaming users are served it is with the requested

throughput. On the other hand, the “Priority Service” strategy

shows the best results for web service, which it was given

more importance, being illustrated the strategies adopted by

the operators when want to give a better service for a specific

group. Hence, it is possible to improve the average network

throughput in 14.3% and the average satisfaction rate in 7.2%,

for users performing Web service when it is used the “Priority

Service” strategy. Hence, this strategy allows better

performances for one specific service, according to operator’s

requirements.

Regarding the average network throughput, it is possible to

notice that this one is also higher when it is applied the

“Carrier 2000 Loading” strategy, but the main difference is in

the satisfaction rate, which presents an increase of 23%

relatively to the “Priority Service” strategy.

V. CONCLUSIONS AND FUTURE WORK

The aim of this work was to analyse the HSDPA and HSUPA

performance in the 900/2000 MHz frequencies, in terms of

traffic management between the two carriers, capacity,

average network throughput and satisfaction rate. This model

was implemented in a simulator written in Matlab, with the

purpose to calculate the statistical parameters in a multiple

users’ scenario with a certain requested throughput, varying

several parameters of each system.

Concluding, the “Priority Service” strategy should only

applied, when the operators intend high requirements for one

specific users’ group, because the better performance of these

users is achieved by reducing one average the throughput and

satisfaction rate of the other services available.

For future work, it would also be interesting to study the

HSDPA and HSUPA 900/2000 MHz with GSM system, in the

same network. Hence, the strategies present in this paper could

be developed with the intention of allowing a traffic

management between the HSDPA and HSUPA 900/2000

MHz cells and 900 MHz GSM cells.

REFERENCES

[1] Holma, H. and Toskala A., WCDMA for UMTS – HSDPA Evolution and

LTE, John Wiley & Sons, Chichester, UK, 2007.

[2] Holma, H. and Toskala A., UMTS 900 Co-Existence with GSM 900, Holma, H., Ahopaa T and Prieur E.

[3] Lopes,J., Performance of UMTS/HSDPA/HSUPA at the Cellular Level, M.Sc. Thesis, IST-UTL, Lisbon, Portugal, 2008.

[4] UMTS Forum White Paper, Deployment of UMTS in 900 MHz band, Oct. 2006.