A new code allocation scheme for UMTS system



Thamer Al-Meshhadany and Khaldoun Al Agha
Laboratoire de Recherche en Informatique (LRI)
Bât 490 Université Paris Sud
91405 Orsay Cedex, FRANCE

phone (33)(0)169156591; fax (33)(0)169156586
Institut National de Recherche en Informatique et en Automatique (INRIA)
Domaine du Voluceau - B.P.105
78153 Le Chesnay

Cedex, FRANCE
email: thamer,alagha  @lri.fr

Abstract—Third generation wireless system is based on the CDMA ac- When evaluating the UMTS mechanism [5] by studying
cess technique. In this technique, all users share the same bandwidth si- points described above, we can observe that firstly, this mecha-
multaneously but with different codes. This sharing generates interfer-
ences that can reduce the system’s capacity when using a weak code al- nism multiplexes all the active user’s services to constitute the
location algorithm. In this work, we analyze the WCDMA capacity as a data flow. This flow is transmitted by using one or several phys-
function of the type and the number of allocated codes. We also study the ical dedicated channels after multiplying every channel by an
way used to assign codes to users. Finally we propose a new scheme that
assigns the code allocation according to the type of the user’s application.
orthogonal code, then we sum these channels and multiply the
Simulation model and results are provided in this paper. result by the scrambling code that is designed by the base sta-
Keywords—WCDMA, control power, resource allocation. tion. The combination of the real time (RT) services and the
non real time services in the same multiplexer can reduce the
I. I NTRODUCTION performance of the RT service transmission (the delay of trans-
mission for the RT services may be augmented).
The data modulation in the UMTS system consists of a se- Secondly, the user transmits in fact two or up to three ser-
quence of two stages: spreading and scrambling. In the first vices simultaneously. The UMTS offers one scrambling code
stage, we spread the user’s information over a bandwidth that and a tree of channelization codes for every user. Consequently,
is large but constant. We replace the 1’s bit (or symbol) of the The code utilization is not optimized because the majority of
user’s information by the chip code and the 1-complement of the OVSF tree codes are unused. Thus, on the basis of the
this chip code if the bit is 0. We provide different data rate by UMTS spreading principles, we propose a new strategy of ser-
replacing each bit with a variable-size chip code in respect to vice management and code allocation that reduces the aggre-
the fixed spreading chip rate [1]. gate number of used scrambling codes in the system. Con-
The Orthogonal Variable Spreading Factor (OVSF) that is sequently, the interference will be moderated and the system
used in this stage provides variable size code and zero cross capacity is improved.
correlation. Every user transmits data on one or multiple chan- Our proposition (called OSSC [6]: One Service, One Scram-
nels according to the information quantity and to the delay re- bling Code) consists of grouping the users that use the same
quired. Thus, every channel can spread the data transmitted service under the same scrambling code by assigning to each
with different codes. This code is called ”the channelization user the adequate channelization code for his active service.
code”. The second stage is the scrambling in which we sum all The base station indicates on the broadcast channel the assign-
the channels of the first stage to constitute one data flow that is ment of this scrambling code. Each user using the announced
multiplied by a unique scrambling code. The scrambling codes service makes synchronization and asks the base station about
are generated from the Gold and Kasami sequences [2], [3], [4] his channelization code. After the assignment of the OVSF
that use pseudo-noise sequences. These sequences are not or- code to this user, he spreads and modulates his data for trans-
thogonal but they provide excellent quality of auto and cross mission.
correlation properties. Thus, It is obvious that the scrambling This work is organized as follow, Section II presents the
codes generate more interference than the orthogonal codes. UMTS and OSSC code allocation schemes; Section III de-
Our analysis for the resource allocation mechanism in the scribes simulation model and provides simulation results.
3rd generation was based on three essential points: how we can
reduce the interference generated from the using of scrambling II. T HE OSSC AND UMTS MODELS
codes in order to increase the system capacity? Secondly, how
we separate the different services taking into account the delay A. The UMTS code allocation scheme
and the Quality of Service (QoS)? Finally, how we simplify the Figure 1 depicts the UMTS code allocation [5] in a cell
algorithm of the resource allocation? where every user can transmit his data in one or several chan-
nels after multiplying each channel by an orthogonal code. We In [6], the OSSC is studied analytically and compared to
sum all these channels to constitute the data flow that is multi- the WCDMA code allocation scheme. The uplink WCDMA
plied by the unique scrambling code assigned to the user by the scheme employs an opposite method than OSSC. It assigns on
base station. scrambling code to each user and then, the user, on this scram-
bling code, multiplex his services by using OVSF tree. These
Cch,1
Cscramb,1 OVSF codes are orthogonal and generate a cross correlation
Different services

equal to zero. Figure 3 illustrate code allocation in WCDMA

Same user

CODING/
MULTIPLEXING and OSSC.
User #1 Cch,K
Cch,1 BS
Different services
Same user

CODING/
MULTIPLEXING
Ms1

User #N Cscramb,N

Cch,K Sc1_(chv+chd)

Fig. 1. UMTS system model.

BS

An example is given in the Figure 3(a) where three users are

Ms3 Sc3_(chv)
actives in the cell. Each user transmits on a unique scrambling Sc2_(chd)
code 
. All services (voice, data or both) are multiplexed on
the scrambling sequence by using separate OVSF codes. Ms2

B. The OSSC code allocation scheme

The OSSC [6] (One Service One Scrambling Code) consists Sci: scrambling code i, i=1,2,3
of allocating a scrambling code for each supported service in chv: voice channelisation code
a cell. The base station indicates on the broadcast channel the chd: data channelisation code
assignment of this scrambling code. Each user using the an-
nounced service makes synchronization and asks the base sta- (a) Code allocation in OSSC
tion about his channelization code. After the assignment of the
OVSF code to this user, he spreads and modulates his data for
transmission. This user processing is illustrated in Figure 2. By
using this mechanism, we multiplex all homogeneous services Ms1
on the same scrambling code and apply the same service re-
quirement on all user flows. Since mixing all users on the same Sc1_ch1 Sc2_ch1
scrambling code, this procedure allows also a best utilization of (voice)
(data)
the scrambling code and reduces consequently the interference
between codes assigned in the given cell. BS

C
ch,1 C
scramb,1 Ms3 Sc1_ch2
(voice) Sc2_ch2
Different users

CODING/
Same services

MULTIPLEXING

(data)
CODING/
MULTIPLEXING Ms2
Scrambling code #1
C
C ch,SF BS
ch,1

CODING/
Different users
Same services

MULTIPLEXING
Sci: scrambling code i, i=1,2
CODING/
chi: channelisation code i, i=1,2
MULTIPLEXING Cscramb,N

Scrambling code #N C
ch,SF
(b) Code allocation in UMTS
Fig. 2. OSSC system model.
Fig. 3. OSSC system model.

Figure 3(b) shows the same example of Figure 3(a) by using

the OSSC scheme: each service has its scrambling code  The study, shown in [6], exhibits the performance optimiza-
and users transmit data or voice by using different OVSF codes. tion of the cell capacity when using an OSSC mechanism. Fig-
ures 4 and 5 illustrate the system capacity in term of the maxi- number of active multi-user.
mum number of users accepted in the system for WCDMA and
OSSC respectively. III. S IMULATION MODEL
Figure 4 exhibits the number of accepted voice users versus The used simulation model consists of a cluster of 64 hexag-
the total number of new users. Several curves are depicted. onal shaped cells where a sample is shown in Figure 6. Edge
Each curve is related to the number of active data user in the effects to handoffs at the cluster boundary are handed by wrap-
cell. The value of the factor of orthogonality is   . ping then around thereby assuming that handoff arrival rates
All curves are superposed until the system capacity reaches its and handoff departures rates from cluster to cluster are equiva-
maximum. The maximum reached depends, of course, on the lent. Among the assumptions made in the study the call dura-
number of active data users. When the active data users are tion is exponential with a mean value of 120s. The cell dwell
high, the number of voice users accepted decreases. time (time a mobile spends in a cell) is also assumed exponen-
tial with mean value 50s. Two types of traffic are considered:
120
voice and data. Voice and data users represent 70% and 30%
100 dt=5 respectively. Arrivals are modeled by Poisson distribution with
dt=10
dt=15 mean 2 . Two 34,576
8 values are considered to distinguish voice
accepted voice‘s users

dt=20
80 dt=25 from data users. Simulation parameters are the same of the an-
dt=30
 dt=35 alytical model in [6]. Table I show values of these parameters.
60 dt=40
dt=45

40

20

0
0 20 40 60 80 100 120
New arrival voice‘s users

  , number of multiple-service users=12, dt: number of

data‘s users
Fig. 4. UMTS anaylitical system capacity.

160

140 dt=10
dt=20
dt=30 Fig. 6. System layout.
120 dt=40
accepted voice‘s users

dt=50
100 dt=60
dt=70
 80
dt=80
dt=90
TABLE I
60
I MPLEMENTATION PARAMETERS .
40

20
item value
0
0 20 40 60 80 100 120 140 160 Bandwidth 3.84Mcps
New arrival voice‘s users voice rate 15kbps
 "!$# %'&(*)+-,./0 1 , number of multiple-service data rate 30kbps
9
users=12, dt: number of data‘s users. 3:,*5;68=<>  ?&A@ B$CED
9
Fig. 5. OSSC analytical system capacity. 3:,*5;68=<F *) ) G CED
Max power transmitted 0.05W
Path loss C0HJI
In Figure 5, same results are illustrated. Comparing to Fig- Noise, K 0 LNMPO:!Q HSR*T
ure 4, the gain obtained by using the OSSC was augmented.
Thus, between 17 and 50 more voice users are accepted de-
pending on the number of data active users. In the same way, The simulation consists in accepting a user while conditions
the number of active data users can be augmented. for a perfect power control is respected. First, the user enters
In short, the gain by using the OSSC scheme varies between in a cell. The power vector of all clients in the cell is recom-
31% and 77%. Other parameters were used to study the be- puted. Users are invited to increase their power because of the
havior of the two schemes. These parameters are: and the additional interference generated by the new user. When the
maximum power is reached, no new users are accepted. The The figure 7(a) exhibits the number of accepted users versus
handoff treatment is similar. Performances are calculated in the total number of new users. Indeed, when the number of
terms of handoff dropping probability and new call blocking users exceeds 2000, the system capacity is obviously increased
probability. All simulations are implemented by OPNET. by 2% for a low system load and 8% for a high system load.
Figures 7(b) and 7.c compare the performance of UMTS
and OSSC under two orthogonal factor values ( VW0 1 and
XY0 G . These figures show clearly a gain of 50% for new call
80000 probability and a gain of 20% of handoff failure probability
70000
when using the OSSC scheme. It obvious that this gain is ob-
tained due to the interface reduced by maximizing orthogonal
60000
allocated codes and minimizing pseudo noise sequences.
Accepted users

50000

U 40000
IV. C ONCLUSIONS
30000 In this work, we have analyzed by simulations under OP-
20000 OSSC,alpha=0.4
NET the OSSC code allocation scheme FDD-WCDMA in the
OSSC,alpha=0.7
UMTS,alpha=0.4
UMTS standard. We focused our study on the interference ef-
10000 UMTS,alpha=0.7
fects on the system capacity. This interference is related to the
0
0 20000 40000 60000 80000 100000 120000 140000 number of scrambling codes in each cell. The essential of our
Total users work separates the effect of the interference into two different
sets: the interference that is generated by orthogonal codes and
(a) Number of accepted users. the interference that is generated by scrambling codes.
Simulation results show potential gain when using one
scrambling code per service. The quantity of interferences is
0.5
reduced on users communication and more users are accepted
0.45
in the cell. Also, probabilities of new call block and handoff
Newcall blocking probability

0.4 failure are reduced.

0.35
R EFERENCES
0.3
[1] TSG RAN WG1, “Spreading and modulation (FDD),” Tech. Rep. TS
0.25 25.213, 3GPP, mars 2000. http://www.3gpp.org.
[2] E. Dahlman, B. Gudmundson, M. Nilsson, and J. Sköld, “UMTS/IMT-
0.2 OSSC,alpha=0.4
OSSC,alpha=0.7 2000 based on Wideband CDMA,” IEEE Communication Magazine,
0.15
UMTS,alpha=0.4 vol. 36, pp. 70–80, Sep. 1998.
UMTS,alpha=0.7
[3] W. M. Prodip Cnaudhury and S. Once, “The 3GPP proposal for IMT-
0.1 2000,” IEEE Communication Magazine, pp. 72–81, Dec. 1999.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
[4] E. H. Dinan and B. Jabbari, “Spreading codes for direct sequence CDMA
Erlang
and Wideband CDMA cellular networks,” IEEE Commiunication maga-
zine, vol. 9, pp. 48–54, Sep. 1998.
(b) New call block probability [5] TSG RAN WG1, “Multiplexing and channel coding,” Tech. Rep. TS
25.212, 3GPP, mars 2000. http://www.3gpp.org.
[6] T. Al-Meshhadany and K. A. Agha, “OSSC: A new scheme for code al-
location in WCDMA,” in IEEE PIMRC2001, (San Diego, USA), sep.-oct.
0.16
2001.
0.14
Handoff failure probability

0.12

0.1

0.08

0.06

0.04 OSSC,alpha=0.4
OSSC,alpha=0.7
UMTS,alpha=0.4
0.02 UMTS,alpha=0.7

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Erlang

(c) Handoff failure probability

Fig. 7. System performance.