Performance Analysis of Various InterferenceCancellation Schemes in CDMA Systems

Himanshu K. Patel Rajendra B. Jadejahkpatel@hkpatel.com rbjadeja200 I1yahoo.com

Institute of Technology, Nirma University. C.U.Shah Engg. College, Saurastra University

Abstract-- In this paper, various interference cancellation final section deals with the conclusions arrived by the varioustechniques that are used in DS-CDMA are analyzed and their schemes for the performance evaluation and suitability for theperformance characteristics are simulated and evaluated. The 3G cellular systems. The various interference cancellationconventional matched filter receiver, decorrelating detector, techniques performance is compared with the DC-PICsuccessive interference canceller and parallel interference (decorrelating detector and parallel interference canceller)canceller are considered. Combination of parallel canceller and . .decorrelating detector is been modeled for performanceconsideration & comparison. For spreading sequence, some II. MATCHED FILTER (CONVENTIONAL DETECTOR)portion of the long m-sequence generated from the primitivepolynomial is used. This sequence doesn't hold good in terms of First, a synchronous CDMA model is presented in [1]. Theproperties and performance. To employ better spreading 'K' numbers of users are active. The users transmit bysequences, the gold sequences generated from the preferred pairs employing the modulation scheme as BPSK. The receivedare also applied for the performance evaluation and comparison. signal will be the sum of antipodal modulated synchronousFor simulation, Synchronous and AWGN (Additive White waveform asGaussian Noise) are assumed. Performance of those interference kcancellation techniques are simulated and the results are y(t) = Akbksk (t) + an(t)),t1E (0,t) (1)highlighted which proves to be very much suited for DS-CDMA k=1systems.

Index Terms-Interference Cancellation, MAI, detector, gold Where T is the bit duration and S is the signature waveform.sequence. The signature waveform is said to be zero outside the interval

I. INTRODUCTION (O,T) so there is no inter symbol interference.

DS-CDMA is one of the most popular and widely used Ak is the amplitude of the received waveform of the Kth user,techniques used in wireless communication. It is chosen due to bk E {-1,+1} is the message bit transmitted by the user.large capacity and immunity to jamming. However theperformance is been limited by multiple access interference T(MAI). This is caused by the cross correlation of users IlSk =JSk (t)2 dt =1 (2)spreading sequences. 0

Although making the sequences, which are spreaded, as Where n(t) is the white Gaussian noise of unit powerorthogonal among each other, could eliminate MAI multipath 2component and asynchronous makes them non-orthogonal. In and thrNg / is th nis powe ther ive saorder to overcome these types of problems many interference passed through a bank of filters as shown in the figure1.cancellation schemes are proposed. To evaluate, severalinterference techniques are reviewed and simulated. In the nextsection a conventional decoding method is introduced. It is a MatchedfiflrYeoptimum receiver incase of single user, but performance Userldegrades when multiple-users are present. In the followingsection the decorrelating detector, which eliminates MAI bymultiplying the inverse of the cross correlation matrix of the Mathed filter Yspreading sequences, is considered [1]. The next section deals User 2with the successive and parallel interference scheme, whicheliminates MAI by subtracting from other interferers,estimated signals. The successive interference canceller iscalled the decision driven multiuser detector. The parallelinterference canceller can be used for several stages[2][7]. The Mthdf~trycombination of this with the decorrelating detector is also Uesimulated for grading the performance in case of multiusers. Inthe last section simulations are reviewed, evaluated and the Fg1 ovninlMthdFle

1-4244-0726-5/06/$20.OO '2006 IEEE 660

III. DE-CORRELATORThis detector is the simple extension of matched filter for one In this section, the decorrelating detector is introduced. Theuser. The performance of the detector is optimal for single user . 'but not for the multiuser environment. In the synchronous case oss orrelan mf weReutptheoutput of the bank of the matched filter by R-1, thethe output of the Kth matched filter is Yi R-1y = R1 (RAb + n)to Yk as (13)

=Ab+R-lnT We observe that the Kth component of R-1 y is free from

yi = fy(t)SI (t)dt interference caused by any other users. The only source ofo (3&4) interference is the background noise. To get the users decodedT bit, we can simply take the sign of the each of the components

Yk = fY(t)Sk (dt of R1y.0 Y

Alternatively, it is stated as of the form b = sgn(R y) (14)Figure 2 depicts the detector of this type. There are several

Yk = jAkbk+ Z AbjPjkfk (5) good properties about this detector. The most important is the

Where performance, which doesn't depend on the signal energy.T There are certain drawbacks like boosting of noise power and

nk = T-n(t)Sk (Odt (6) calculation of RK' which is difficult in practicaljxik (0d implementation.0

It is convenient to represent the output of the matched filter inmatrix notation. To represent we employ the cross correlation I

. . ~~~~~~~~~~~~~~~~ ~ ~~~~~~~~~ldlhedhlWr;I-matrix is defined as uC 1

P11 P12 P..Plk Mth

R P21 P21 ..P21 7 t YR = (7)

_Pkl Pk2 . Pkk ;Where

T - drKIIIfxSI (t)S (t) dt, wherei , j I = .. .......... k (8)0

Fig 2. Decorrelating Detector

It is convenient to state the output of the matched filter asy = RAb + n (9) IV. SERIAL INTERFERENCE CANCELLER

Where SIC is one of the decision-driven multiuser detectors. Let usr=T consider the two user synchronous case, it is demodulated by

Y = LY' Y2 *L Yk] its matched filter as

b = [b1 b2 bk ]T (10)

A=diag{A,,A2..., Ak} b2 =sgn(y2) (15)

Demodulating the signal of the user2 with b2, we obtainand n is the zero mean Gaussian random vector with A2b2S2 (t). The estimated signal is subtracted from thecovariance matrix equal to

T 2 received signal, which yieldsE nn = a R (11)To get the users decoded bit, we simply take the sign of each y(t) = y(t) - A2 b2S2 ()of the components as the outputs. A

b = sgn(y) (12)

661

considered. Without loss of generality, the error rate of a firstuser is plotted. Figure 5 shows the conventional filter response

r ~~~~ofgold sequence respectively. When employing rn-sequencethe performance degrades as the number of user increases. It is

the matched filter receiver is not optimal when multiple usersYM ~~~~~~~~~~arepresent. Its performance improves when the

A2 crosscorrelation becomes small as in figure 5. The gold codesdoes this since it has good crosscorrleation properties.

PERFORMANCE OF MATCHED FILTER(GOLD CODE)r ; ; -A-- ~ - 20 users

--e-10

usersFig 3. SIC (2 user case) ------users---------

In the signal y(t) the interference from user2 is removed. This 16 single-usernew signal y is, now processed with the matched filter for sl1.----------------So we obtain the decision as T

=sgn(y1-A2psgn(y2))(17)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ---- ----

=sgn(A1b1+A2(b2-b2)p+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ---n,S----1>1------- ------L --L -------I -LWhere <n,sl> is thecross correlation between n and si. The 2~~~~~~~~~~~~~~~~~~~~~~~L-- L--- -L- -!---

usersuccessivecancellationdetector is shown above in the~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~figure. In K Ausr ucesiv cancellation, the users interference 1- T--- T --componentissubtracted one by one at a time. When making a12 3 4 5 6 7 6 9 16 -- - --r r-- -- T -- T

decsion(yabout Kt user,ywe7assum thttedeiin of ses ELN IdE--K± 1, K are correctand we eliminate the presence of the users~~~~~~~--- - ::- --

detector structure is shown in the figure4 below. -- - -e- 6 users~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~rI-------I-- sigl user

figure. In K use sucssive 10clainh uesitreec

compoent i subracte one y on at atime.Vvlhn maknga12 3 4 6 6 89 1

K+11arecorretVIn SeeimULatIONSpeeneofte sr Fig 65 Performance ofdetcredflating Deecor(old sequec)Gol codes] ar eeae rmpeere ar f<5 n<7>wicVa. ue oPAALLIThREcopRison.ThNELEperiodNEofEtheATR(LCDE

Incoderist31ForteSImlCit, SynPChronecou standaWGNal Ac anne is---------

662 -------------L--------------- 2 s r

PERFORMANCE OF SIC(GOLDCODE)_ uesPERFORMANCE OF PIC(GOL0 CODE &0C AS FIRST STAGE5 ses20 users

-- -- -- --:- -A,-A rs--2- users- -- ----------1--- --users1--------------- ------10-users--iE- 5 users

-- -- -- in le us r --sin--gle-- -user---- --sin--- --le--userg e se

-L

10 -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-----------1--------------

D~~~~~~~~ .2 [i~~~~~~~~~~~~~~~~~~~~-----LU 1::: :::::::::: 10::]::: :Lu Lu~~~~~~----------- ----------- -

1 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~10~~~~~~~~~~~~~~~~~~~~~~~~r --------. -----

12 3 ~ ~ ~ ~ ~ ~~~~~4-----------------r - - 5 ---------I 10 1 23 4 - - - -6 --

Eb/N--udE----- Eb/N udE ------------

Feg 7 Pefomac o SC gldse)-e 9 Comb-ination-of PlC & Decorrelator (gold seq)

PERFORMANCE OF PIC(GOLD CODE) VII. CONCLUSION~~~~~~~~~---------A------ - 20 users ----- -----------L

~~~~~2 M~-+0uer h vaiu nefrne aclaintcniusaervee6- ----5users an siuae.Temthdfltr hc sotmmol

m -- -in--gleusrwe-in l s ri present, isno nce sail-otiu fo the101 L ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~mult-usercase. Otermlt-sedtctr ar mor effective.

--------------- r- ---By---using--r- gold-- -codes -significant --performance--rimprovement----- --- can---------------- ---- :------be---obtained----as ---it-holds good ---cross---cor lto propertes-whe

L -------- -------compared-----to the i -sequences. Finally-- emplying good code---------plays---------- -a----vital---- role--- for--- eliminating---- the----comp nents

-----apait-icresean Syte-prfomace

~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ II.REEENE----[1]-----.---Verdu-------"Mult--iuser--- Detection,"---------- Cambridge---- University-- Press, 1998

1 [2]---------5.------Moshav------i,-----"Multiuser- Detection- for----DS-CDMA--- Com uniatin"-EE---------------------------------Communication-----Magazine, ---pp. --124-136, ---Oct.1996.

----------------------------- -------[3]---.-B.--Wicker,--"Error---Control--Systems---for-Digital---Communications--

1 2 3 4 5 6 1 Hue1998. 5 6

Fig.8.Performance ofPIC (gold seq) CDMA, IEE trabnsactions onPCommunicoreations (gold4 ppe28-68

considered. Heetedcreltrisdas the fierstsaend Vl8no,p.6390thePl -as-th second--- stage.--The---------fiur 9srshowsosinefrectheltintchius r eveeperorane pot- wicproves----------to-----be--moreiulaeeffective, bytr,whih s doing nleffctvey hathe-- matched-filter-------- response- as ushow innsnleue thesn, sntneesrlyotmmoh

figures 5. The plots in Figure 6 shows the performance of~~~~~~~~~~~~mutiusr ase 0ermui-se eec rsar mree c ve

performance of SIC & PlC.- In-- case-------ofusngglgold seqfcat themac iprvmetaperformance of PlC iscompa--------rabl- t tha of SIC.----

r-- r -6 63--