2009 IEEE International Advance Computing Conference (IACC 2009)Patiala, India, 6-7 March 2009

Vacant Code Precedence and Call Blocking Reduction in WCDMA Systemsbased on OVSF Codes

Davinder S. Saini, Vikas Hastir, Munish SoodDepartment ofElectronics and Communication Engineering

Jaypee University ofInformation TechnologyWaknaghat, Distt. Solan.(HP.)

davinder.saini@iuit.ac. in, vikas.hasthir@juit.ac. in, munish.sood@iuit.ac. in,

Abstract: In this paper a code assignment schemecalled Next Code Precedence High (NCPH) is SC1,Ch _ UE,proposedfor UTRA-FDD (Universal Terrestrial Radio Node Bi__Access Frequency Division Duplex) systems based onOVSF (Orthogonal Variable Spreading Factor) SC- Ch2channelization codes. Using the proposed scheme, thenumber of codes searched before finding suitable [ UE2vacant code is least. Also the external code Lifragmentation which can lead to code blocking is lessbecause of the compact nature of code assignment SCi, Ch3 UE3scheme. Simulation results are presented to compare Hthe reduction in call blocking probability of theproposed scheme with the existing novel assignment (a) Forward Linkschemes.

SC1, Ch1Keywords: WCDMA, OVSF codes, spreading factor

S,C

(SF), code blocking, code (channel) assignment and Nnode Bireassignment

1. Introduction SC2, /UTRA-FDD/WCDMA [1,2] system differs from UE2existing 2G and 2.5G systems in the use of directsequence spread spectrum multiple access method. In LiWCDMA, the channels in the forward link and reverse C3, Ch3link use two set of codes known as scrambling codes UE3and channelization codes. In forward link, scramblingcodes are used for interference mitigation while the LIchannelization codes are used for rate matching anduser separation. In the reverse link, scrambling codes (b) Reverse linkare used for user identification and interferencemtai whil chan o ce ae ud f rate .........................Fig.1. Scrambling and Channelization code usage in WCDMA.

Figure l(a) shows the forward link transmission from th nodematching. The use of these codes is illustrated in Fig. 1. (Node B,) to three UEs. All the channels use same scrambling codeThe channelization codes used in WCDMAI4 are OVSF and different channelization code. Figure l(b) shows reverse linkcodes. The code generation in OVSF code tree is given link transmission from three UEs to Node Bi. All channels/usersin [3].The spreading factor of 0VSF codes varies as uses different scrambling codes. Channelization codes need not to

be different.

TABLE I CODE ASSIGNMENT SCHEMES

S.No. Type T References

[1. | Single Code Assignment without time Random Assignment [4], Leftmost Code Assignment [4],. |multiplexing and without reassignments |Fixed Set Partitioning [5], Adaptive Assignment [7], Ours|

Single Code Assignment without time TDynamic Code Assignment [6]____multiplexing and with reassignments ________________________

. Singlge Code Assignment with time T NnBoknOS oe 8

4 |Multiple Code assignment scheme for T Mlilrk obnr 9[4 1 quantized rates Mltil raeCminrl9

978-1-4244-1888-6/08/$25.00 Q 2008 IEEE 812

4,8.,,512 in the forward link and 4,8....,256 in the the layer l is 21-'R. There are 8 different arrival classesreverse link. The possible data rates in the forward link of users with data rate R, 2R., 27R. When the neware R,2R....128R (where R is 7.5kbps). The total chip call arrives the vacant code is assigned to it. All of itsrate transmitted through channel is 3.84Mcps. Due to ancestors and descendants are blocked as discussedthe orthogonality condition in OVSF code tree, only earlier. In addition, the code next to the assigned codeone code can be used from root to leaf. This gives code as well as all of its ancestors and descendants are givenblocking in OVSF codes resulting in new call blocking. a two dimensional precedence number (x,y), where 'x'Therefore the spectral efficiency and throughput is the layer number and 'y' is the vacant code prioritydecreases if code blocking problem is kept unsolved. number. The precedence number is to be used by nextA number of OVSF code assignment and Codes Blocked

reassignment strategies are proposed to remove codeblocking. In general, the strategies can be put in two Descendantscategories named single code assignment strategies and Cl l,2n-1, C,-1,2n in layer /-Imulti code assignment strategies as shown in Table I. C-2,4n-3,----, C,-2,4n in layer 1-2The single code assignment schemes use one coderequiring single rake combiner in the UE (User laEquipment). The multiple code assignment schemes cl,2/-'n-2±1.-1,21-''n ayer 1use multiple codes with the requirement of multiple Ancestorsrake combiners in the UE. In random assignment (RA) c n in layer 1+1scheme [4], the incoming user picks the code from '2appropriate layer randomly. In leftmost code c n in layer 1+1assignment scheme (LCA) [4] code usage in the code /+2, 41tree starts from left. In the fixed set partitioning (FSP) .....................................

scheme [5], the code tree is divided into a number of c in layer 8sub trees according to the input traffic distribution. RA, 8.028llLCA and FSP schemes suffer from the limitation oflarge blocking probability and smaller throughput. Codes with Precedence number

In dynamic code assignment (DCA) [6] scheme, theblocking probability is reduced using reassignments C,n C1,n+1 inlayerlbased on the cost function. The DCA scheme requires C1-1,2n+,l Cl,2n+2 in layer l-extra information to be transmitted to inform the C1-2,4n+, C1-2,4n+4 in layer l-2receiver about code reassignments. The adaptive C..................in.layer....assignment scheme (ADA) [7] does code reservation c 1l. c 1 - in layer 8adaptive to the distribution of input traffic classes. The 1,2 n+1 '.1,2 n+2assignment scheme based on NOVSF codes [8] makeblocking zero, but require large number of orthogonal Fig.2. Illustration of blocked codes and the codescodes. Also time multiplexing may be required which which are assigned precedence number correspondingadds to the complexity of the system. The assignment to busy code C,,nscheme [9] uses multiple rake combiners (multiplecodes) to avoid code blocking. The performancecomparison of some of these schemes is given in [10].The remainder of the paper is organized as follows, incoming calls. Let code C,,n is occupied by the new

Section 2 explains the proposed NCPH code user. The codes blocked and the codes which are givenassignment scheme. Section 3 presents simulation precedence number are given in Fig.2.model and results. The paper is concluded in section 4. The vacant codes are listed according to the

precedence number used. The list is prepared for every2. Next Code Precedence High (NCPH) Code layer and updated for every new call. If two more or

Assignment Scheme more codes in a layer have same vacant code priorityConsider an OVSF CDMA system with 8 layers. The number, any of them can be used for assignment. Whencode in layer l, le [1,8] is represented by C.,, where n a code is used for assignment, precedence numbersvaries from 1, 2.,281. The maximum capacity of (priority numbers) of codes and all of its ancestors andeach layer and the system is 512R (3.84Mcps). We descendants is given arbitrary large value so that theyintentionally use SFs 27, 26........,2° instead of 29,28......22 are not candidates for code assignment procedure.(512,256,... .4) for mathematical simplicity. The When the call iS completed for a code, itS prioritynumber of codes in layer l is 28-l. The spreading factor number iS made higher than the highest vacant codein layer l is 28-/. The data rate handled by the code in priority number. So the code precedence number iS

2009 IEEE Internlationlal Advance Computing Conference (IACC 2009) 813

modified for every code assignment and call * When the call using code Cl, is completed, itscompletion. The precedence number has following ancestors and descendants are given precedenceproperties number higher than the highest precedence* Initially when the code tree is vacant, the Code Cl,n number in the code layer (e.g. 129 {27+1 } for layer

is represented by two dimensional precedence 1, 65 {26±+1 }for layer2 and so on.)number (I,n). * If two or more codes have same (I,n) number, any

Layer 5,k 5,k 5,k

54,k 4,k 4,k 4,k

4

34k 3 k 3,k 31 3k k 3,k

3

2, 2,1 2k 21 2,, ,2k 2,1 2 ,k 2,

2Ut /\ 0 0 0 0 0 0 ,1\, ~~~tIkAki \0 0 t{k Il,tki\ \

l, ,k1,1,lkl,k l,k l,k ', ,91212l,k l,k l,k l,k l, lk 1,2 1,2 1,2 1,2(a) (b) (c)

Busy Code Blocked Code 9 Vacant Code Vacant Code with Precedence number

k=large positive value (say 10000), signifying non vacant code

Fig.3. Illustration of Next code Precedence High code assignment scheme. Fig.2.a shows code assignment,blocking and precedence number assignment for first 2R rate call arrival. Fig.2.b and c illustrates theassignment, blocking and precedence number assignment for second (Rate 4R) call and third (Rate 8R) call.

* When the code is occupied by the call, its of them can be used for new call.ancestors and descendants are given arbitrary large The assignment scheme is illustrated in Fig.3. Thevalue ofpriority number say 10000, signifying that vacant code tree with five layers is considered. Forthe codes are in use. simplicity we did not consider the precedence number

of the vacant codes. Let the three calls with rate 2R, 4REnter arrival rate, service and 4R arrives in sequence. The code assignment and

time, data rates precedence ordering for three calls is shown in Fig.3.a,-hT b and c.

| Generate new callwith rate R' 3. Simulation parameters and results

3.1. Input DataYe * Call arrival process is Poisson with mean arrival

(Capacity used + R > Discard call rate i 1 -128 calls/ unit time.Max_Capacity

* Call duration is exponentially distributed with aNo mean value 1/, =1 units of time.

* Possible OVSF code rates considered are R, 2R,Pick the vacant code withthe highest precedence 4R and 8R corresponding to four different arrival

classes.* Total number of codes in layer 4 is 16 making

maximum capacity of the system 128R.Block ancestors and

descendants. 3.2. Results

The Flowchart for the simulation model is shown infl r Fig.4. Simulation results are demonstrated to show the

Change the precedence of reduction in call blocking probability in NCPHcodes according to the new scheme. The call blocking in NCPH scheme is

busv codecompared with the call blocking of Random

Fig.4. Flowchart for simulation model

814 2009 IEEE Internlationla/Advance Computing Conference (IACC 2009)

Blocking probability for rate distribution [0.25,0.25,0.25,0.25] rates (higher data rates) (Fig.6.) dominates the traffic50-

NCPH load and in the third case real time users (lower data40 FSP .1'ERA rates) (Fig.7.) dominates the traffic. The simulation is

DCA done for 5000 users and result is the average of 10me ;- - XFX-:t simulations. Simulation results shows that the call

30 blocking in NCPH code assignment scheme is lessD , !, compared to FSP, LCA and RA assignment schemes

' 20 and is more than DCA scheme.0

10 - 4. ConclusionThe OVSF codes are the limited resources in the

2 forward link of WCDM4. The proposed NCPH schemeo 20 40 60 80 100 120 140 provides an efficient way to use OVSF codes. The code

Traffic load tree is less fragmented due to the compact nature ofFig.5. Blocking probability vs traffic load

code assignment scheme. It provides reduction in codeblocking and hence new call blocking leading to higher

Blocking probability for rate distribution [0.1,0.1,0.4,0.4] system throughput and higher code utilization. The UF609\ <NCPH complexity need to be analyzed due to storage of

50 LCA vacant code precedence numbers.~~~~~~~~~~~FSCP, 40 , X References

M / X [1] F.Adachi, M. Sawahashi, and H.Suda, "Wideband CDMA for2 30 next generation mobile communication Systems", IEEE

Communication Magazine, vol. 36, pp. 56-69, Sept.1998., 20 / . [2] E. Dahimnan, B. Gudmundson, M.Nilsson, and J. Skold,m° / 1k=; < 3 "UMTS/IMT-2000 based on wideband CDMA", IEEE

10 ' tt * A \e X 7 t-Communication Magazine, vol.36, pp. 70-80, Sept, 1998.10 X:- S[3] F.Adachi, M.Sawahashi, and K.Okawa, "Tree structured

generation of orthogonal spreading codes with different lengths0 S t; X --for forward link of DS-CDMA mobile radio", IEEE Electronic

0 20 40 60 80 100 120 140 Letters, vol. 33, pp.27-28, Jan. 1997.

Fig.6. Blocking probability vs traffic load [4] C.M.Chao, Y.C.Tseng and L.C.Wang, "Reducing Internal andExternal Fragmentation of OVSF Codes in WCDMA SystemsWith Multiple Codes", IEEE Transactions on WirelessCommunication, vol.4., page 15 16-1l526, July 2005.

Blocking probability for rate distribution [0.4,0.4,0.1,0.1]Comnctn,vl4 pg15652 Jy20570 [5] J.S.Park and D.C.Lee, "Enhanced fixed and dynamic code

assignment policies for OVSF-CDMA systems", in proc. of60 ICWN 2003, Las Vegas, June 2003.

[6] T.Minn and Kai-Yeung Siu, "Dynamic Assignment of orthogonal50 DCA variable spreading factor codes in W-CDMA", IEEE J.

___40_/ LCA Selected Areas in Communication, vol. 18, no.8, pp. 1429-~~~ 40 ~~~~~RA 1440, Aug. 1998.

M ~~~~FSP "dpieshm o-O NCPH [7] D.S.Saini and S.V.Bhooshan, "Adaptive assignment scheme for

30 OVSF codes in WCDMA", Proceedings of IEEE ICWMC,Bucharest, July 2006.

m° 20 / X [8] H.Qam, "Non-blocking OVSF codes and enhancing networkm1 capacity for 3G wireless and beyond systems", Elsevier

10 I Journal ofComputer Communication, pp. 1907-1917, 2003.[9] Li-Hsing Yen, Ming-Chun Tsou, "An OVSF code assignment

0 X X scheme utilizing multiple RAKE combiners for W-CDMA",0 20 40 60 80 100 120 140 Elsevier Journal of Computer Communications 27(16) 1617-

Traffic load 1623 (2004).Fig.7.Blocking probability vs trafficload [10] D.S.Saini and S.V.Bhooshan, "Assignment and Reassignment

Assignment (RA), Leftmost Code Assignment (LCA), Schemes for OVSF Codes in WCDMA", Proceedings ofACMAssignment (RA),Leftmost Code Assignment (LCA), international conference IWCMC, Vancouver, pp. 497-502,

Fixed Set Partitioning (FSP) and Dynamic Code July 2006.Assignment (DCA) schemes discussed earlier. Theprobabilities for different class users are denoted by(P",p2,p3,p4), wheref,P]P2,P3 andp4 are the probabilitiesof arrival rate R,2R,4R and 8R users. We considerthree different distributions of arrival classes. In thefirst case traffic load is uniform (Fig.5.) for the fourarrival classes. In the second case, non-real time traffic

2009 IEEE Internlationlal Advance Computing Conference (IACC 2009) 815