B.Sathish Kumar 1, K.R.Shankar Kumar 2 1,2

Transcription

1 A ove Approach of Linear Parae nterference Canceation in Upin OFDMA Systems with Carrier Freuency Offsets B.Sathish umar 1,.R.Shanar umar 1, Department of Eectronics Communication Engineering Sri Ramarishna Engineering Coege, Coimbatore. Taminadu, ndia. E-mai : 1 sathish_mtp@yahoo.com, shanwire@gmai.com Abstract n upin orthogona freuency division mutipe access (OFDMA) systems, muti-user interference (MU) occur due to different carrier freuency offsets (CFO) of different users at the receiver. ere, we present a mutistage inear parae interference canceation (LPC) approach to mitigate the effect of this MU in upin OFDMA. The proposed scheme first performs CFO compensation (in time-domain) foowed by DFT operations (where is the number of users) mutistage LPC on these DFT outputs. We scae the MU estimates by weights before canceation imize these weights by maximizing the average signa-to-interference ratio (SR) at the output of the different stages of the LPC. We derive cosed-form expressions for these imum weights. The proposed LPC scheme is shown to effectivey cance the MU caused by the other user CFOs in upin OFDMA. We show that our approach performs better than the existing techniue by comparing the bit error rates of the existing system our proposed system. eywords Linear Parae nterference Canceation (LPC), Weighted Linear Parae nterference Canceation (WLPC), Carrier Freuency Offset (CFO), Mutipe-user nterference (MU).. ntroduction The principes of orthogona freuency division mutipexing (OFDM) moduation have been around for severa decades. owever in recent years, this technoogy has uicy moved out of the academia word into the rea word of modern communication systems. OFDM is a muti-carrier moduation scheme that encodes data on to a radio freuency (RF) signa. Unie conventiona singe-carrier moduation schemes such as AM/FM (ampitude or freuency moduation) that send ony one signa at a time using one radio freuency, OFDM sends mutipe high-speed signas concurrenty on speciay computed, orthogona carrier freuencies. The resut is much more efficient use of bwidth as we as robust communications during noise other interferences. OFDM [1] in its primary form is considered as a digita moduation techniue, not a muti-user channe access techniue, since it is utiized for transferring one bit stream over one communication channe using one seuence of OFDM symbos. owever, OFDM can be combined with mutipe access using time, freuency or coding separation of the users. n Orthogona Freuency Division Mutipe Access (OFDMA), freuency-division mutipe access is achieved by assigning different OFDM sub-channes to different users. OFDMA supports differentiated uaity-of-service by assigning different number of sub-carriers to different users in a simiar fashion as in CDMA [1], thus compex pacet scheduing or media access contro schemes can be avoided. Orthogona Freuency-Division Mutipe Access (OFDMA) is a muti-user version of the popuar Orthogona freuency-division mutipexing (OFDM) digita moduation scheme. Mutipe access 8 nternationa Journa of Computer Science Technoogy is achieved in OFDMA by assigning subsets of subcarriers to individua users. This aows simutaneous ow data rate transmission from severa users. Orthogona Freuency Division Mutipe Access (OFDMA) [9]-[10] assigns a subset of subcarriers to individua users the transmission is simutaneous, as shown in Fig.1. Each OFDMA user transmits symbos using subcarriers that remain orthogona to those of other users. More than one subcarrier can be assigned to support high rate appications. Simutaneous transmission from severa users resuts in better spectra efficiency. Fig. 1. OFDMA- Mutipe Access. OFDMA OFDMA can be seen as an aternative to combining OFDM with time division mutipe access (TDMA) or time-domain statistica mutipexing, i.e. pacet mode communication. Lowdata-rate users can send continuousy with ow transmission power instead of using a "pused" high-power carrier. Constant deay, shorter deay, can be achieved. OFDMA is considered as highy suitabe for broadb wireess networs, due to advantages incuding scaabiity MMO-friendiness, abiity to tae advantage of channe freuency seectivity. Muticarrier moduation schemes, as in OFDMA, eiminate or minimize inter-symbo interference (S) by maing the symbo time arge enough so that the channe-induced deays deay spread being a good measure of this in wireess channes are an insignificant fraction of the symbo duration. Therefore, in high-data-rate systems in which the symbo duration is sma, being inversey proportiona to the data rate, spitting the data stream into many parae streams increases the symbo duration of each stream such that the deay spread is ony a sma fraction of the symbo duration. Mutipe-access schemes are orthogona when an idea receiver can competey reject arbitrariy strong unwanted signas using different basis functions than the desired signa. CFO refers to the difference in carrier freuency at transmitter receiver. The transmitter woud transmit at the nomina carrier freuency. At the Receiver the unmoduated freuency is reuired for reception; is usuay generated in the Receiver with crystas or so. owever, it is not physicay possibe to mae the Receiver freuency exacty match the Transmitter freuency. This offset is termed CFO. A time interva eft vacant (i.e., during which no data is sent) on a transmission channe that can be used for synchronization /or compensating for a signa distortion is nown as guard interva. To eiminate S amost competey, a guard time is introduced for each symbo.

2 n the upin, each users signa may be distorted by an independent carrier freuency offset (CFO)[],[6],[8]-[10], which impairs the orthogonaity of the subcarrier signas, if not propery compensated, resuts mutiuser interference (MU) aong with significant performance degradations. The presence of a carrier freuency offset (CFO) introduces not ony C (inter-carrier interference) S but aso mutipe access interference (MA) between users in the OFDMA upin. Mutipe-access schemes are said to be orthogona when an idea receiver can competey reject arbitrariy strong unwanted signas using different basis functions than the desired signa.. deaized System Mode An upin OFDMA system is considered with users, where each user communicates with a base station through an independent mutipath channe as shown in Fig.. E hn, = E hnq, = 1/L with zero mean, where h i n,, h n Q are the rea imaginary parts of h n. The channe coefficient in freuency-domain is given by jπ n h e, L 1 = n n= 0 E = 1 The received baseb signa after coarse carrier freuency tracing (eaving some residua carrier freuency offset) is given by () jπ n n n n, g i= 1 r = s e + z n 1 (4) Fig.. Upin OFDMA System Mode t is assumed that there are subcarriers in each OFDM symbo one subcarrier can be aocated to ony one user. The information symbo for the ith user on the th subcarrier is X i denoted by, where Si is the set of subcarriers assigned Ε X = 1 to user i. ε where i, i = 1,..., denotes ith user s residua carrier freuency offset (CFO) normaized by the subcarrier spacing, Zn is the AWG with zero mean variance. We assume that ε a users are time synchronized i, i =1,.,., is nown at the receiver. Fig. shows the receiver baseb processing incuding CFO compensation is in time-domain guard time remova, - DFT operations (one for each user), Linear parae interference canceation (LPC) in mutipe stages. Then, i = 1 S i = {0,1,,..., 1} S S, i j = φ for i j The ength of the guard interva added is g sampes is assumed to be onger than the maximum channe deay spread. After DFT processing guard interva insertion at the transmitter, the time-domain seuence of the ith user, is given by x n 1 jπ n xn = X e, g n 1 s i (1) The ith user s signa at the receiver input, after passing through the channe, is given by sn = x n h n () h where * denotes inear convoution n is the ith user s h n is non- channe impuse response. t is assumed that zero ony for n = 0,..., L 1, where L is the maximum channe deay spread, that a users channes are statisticay independent. An assumption is made that h n are i.i.d. compex Gaussian Fig..Receiver baseb processing CFO compensation mutistage interference canceation t shoud be noted that the CFO compensation is carried out jπε i r in time-domain by mutipying n with e, i = 1,, (this method of CFO compensation is referred to as the direct method). The received signa after CFO compensation guard time remova for the ith user is given by jπ n yn = rn e, 0 n 1 (5) nternationa Journa of Computer Science Technoogy 9

3 which forms the input to the ith DFT boc. The output of the DFT boc for the ith user on the th subcarrier is then given by,( ) ( ) ( ) = X + X + Ζ, i 1 1, sin j i π + δ π + δ i i = e π sin ( + δi ) δ i is the difference between the ith user th user CFO vaues, given by δ i = ε ε i The channe coefficient component Z is given by (7) (8) (6) is given by () the noise j πn( + ε ) 1 i i Z = Ζ e, n= 0 n (9) t shoud be noted that the nd term in (6) represents the CFOinduced MU present at the DFT output. n the case of singe user detection (SUD), the DFT outputs, s, can be directy used to mae the symbo decision. Additiona processing may be performed on s in order to mitigate the effect of MU. For exampe, mutistage interference canceation techniues can be empoyed to improve performance. n the next chapter, a mutistage weighted inear parae interference canceation scheme which operates on the DFT outputs is proposed. V. Proposed WLPC Scheme t is noted that, for the desired user i, the nd term in (6) represents the CFO-induced MU (i.e., interference from other users = 1,,,, i ) present at the DFT output. Aso, from Euations (6), (7) (8), it can be noted that the amount of this MU depends onδ i s, the difference between desired user CFO ( ε i, ) other user CFOs ( ε, ). The aim here is to cance this nd term in (6) using a mutistage inear PC approach, where an estimate of the MU in a given stage is obtained using the soft vaues of previous stage outputs (without any non-inear operation, e.g., hard decision, on the previous stage outputs). Further, it is nown that the MU estimates in an LPC approach can become uite inaccurate under poor channe conditions (e.g., ow SR, high interference) to such an extent that it may be better not to do canceation. Such situations can be aeviated by scaing the MU estimates by weights (preferaby by some imum weights) before canceation. ere, such a weighted 40 nternationa Journa of Computer Science Technoogy LPC (WLPC) scheme for the upin OFDMA is presented. The proposed mutistage WLPC scheme is shown in Fig.. Let m denote the stage index. The DFT outputs, in (6) are taen as the first stage (m = 1) outputs of the receiver, i.e. =, (1),. n the case of SUD, the symbo decisions are, ( 1 ) made directy from. LPC is performed in the subseuent stages. n a given LPC stage m, m > 1, an estimate of the MU is made based on the soft vaues of the previous stage outputs. These MU estimates are scaed by weights subtracted, ( 1 ) (i.e., canceed) from the DFT outputs, As mentioned earier, the nd term in (6) is seeed to cance. Towards that end, the foowing operation is considered on the, ( 1 ) other user DFT outputs, t is seeed to cance the second term in DFT boc output. ence the foowing operation on other user DFT outputs is considered,,( ) ( ),(1) i Using (6) in the above, it can be written (10),( ) ( ),( ) ( ) ( ) ( ),( j) ( j) ( j) ( ),(1) = X + r r Xr + Ζ = 1 s = 1 s j = 1r s i i i j,( ) ( ) ( ) = X i,( ) ( ),( j) ( j) ( j) + r r X r = 1 s j = 1r s j i j i,( ) ( ) + Z i ( ), ( 1 ) (11) t shoud be noted that the 1st term T1 in (11) is the same as the MU term (i.e., the nd term) in (6), which we wanted to cance. ence, (10) can be viewed as an MU estimate for the nd stage of the LPC, which when canceed (i.e., subtracted) from (6) wi competey remove the MU term (i.e., the nd term) in (6). n the process, additiona terms T (interference) T (noise), which were not there in the 1st stage output, get introduced in the nd stage output. The interference term T introduced in the nd stage can be canceed in the rd stage ( ),() using a simiar MU estimate obtained from, so on. Accordingy, in the proposed WLPC, the interference canceed output of the ith user on the th subcarrier in the mth stage,

4 ,( m), m > 1, can be written as,( ) ( ),( m) =,(1),( m),( m 1) i (1),( ) ( ),,() = X 1,() + + i (14), ( 1 ) is the 1st stage output given by (6),,( m ),( ) is given by (7), is the weight with which the MU estimate is scaed. The WLPC scheme with the weights on a subcarriers of a,( m) 1 users to be unity = is caed as conventiona LPC (CLPC) scheme. n the CLPC scheme, the operations needed for the choice of imum weights MU scaing with these weights are avoided (because of unity weights). owever, performance better than that of the CLPC can be achieved by using imum weights. A proposa to obtain the imum,,( m), i = 1,...,, S weights for the mth stage, i, by maximizing the corres- ponding average SR at the mth stage output is made. n an uncoded system, the symbo decision for the ith user on the th subcarrier at the output of the mth stage can be made,( m ) based on the output. For exampe, the symbo decision at the mth stage output for the case of BPS moduation can be obtained as X,( m) = sgn Re,( m) (1) n the above euation henceforth, overine is used to denote conjugate operation, i.e., denotes the conjugate i of. V. Euations SR Anaysis Optimum Weight Derivation n this section, expressions for the average SR at the output of the nd rd stages of the proposed weighted LPC scheme are derived. Aso, the derived average SR expressions are used to obtain cosed-form expressions for the imum weights,,( m ). t is noted that the average output SR on a given subcarrier wi depend on severa things incuding number of users, channe impuse response, number of subcarriers, CFO vaues, type of subcarrier aocation. ere, two types of subcarrier aocation are considered, namey, i) boc aocation ii) intereaved aocation. n boc aocation, a consecutive boc of subcarriers is aotted to one user, the next boc to another user, so on. n intereaved aocation, the subcarriers of each user are uniformy intereaved with the subcarriers assigned to the other users. 5.1 Average SR at the nd stage output From (1), the weighted interference canceed output of the nd stage (m = ) for the ith user on the th subcarrier is given by ( 1 ) ( ) ( ),( ),( u) ( u),( ) = X p,(),() v 1 p p s p u 1 u su v = ip u i,,( ) ( ) Ν = Ζ,() Z i i (15) (16) The terms in (15) (16) represent the interference noise terms introduced due to imperfect canceation in using the soft output vaues from the first stage. After coherent combining using, the fina output is given by 1 i (17),( ) ( ),,() = X,() + + = = ( ) p Considering the factor in, it is noted that the channe coefficients on subcarriers of different users i1 i, ( ) 1 i, 1 Si1, Si, i1 i, are uncorreated because a users channes are assumed to be independent. owever, from (1), it can be seen that the channe coefficients ( i) on different subcarriers of the same user i, 1, 1, Si, are correated. Aso, this correation depends on the subcarrier aocation. ing the correation between ( ) 1 i in the SR anaysis is tedious. Therefore, to ( ) simpify the anaysis, it is assumed that 1 i are uncorreated. Accordingy, the variance of be obtained as = i ( 1 ),( ),( u) ( u),( ) = p,(),() 1 p s u 1 u su v v = p u i, (19), can (18) nternationa Journa of Computer Science Technoogy 41

5 And the variance of,, can be obtained as = i,, (0),( ) = + (,()) + i,( ),( u),( u),( ) ( ), η Re η,() 1 u 1 v v,() s = u su 1 s = i u i, i (1) jπn ε εi + ( ), ( ) 1 η = E Z Z e = n= 0 () The average SR on the th subcarrier of the ith user at the SR,() output of the nd stage, can then be obtained as SR,() =,( ) ( ),,() i 1 + () 5. Average SR at the nd stage output The soft vaues of the interference canceed nd stage outputs,, of different users, i S, are used to reconstruct (estimate) the MU on the th subcarrier of the desired user i in the rd stage. These MU estimates are then scaed by,() canceed. Accordingy, the weighted interference canceed output of the rd stage for the ith user on the th subcarrier,(), is given by,() = X F + +,( ) ( ), ( ) F = 1 w,() ( (1 w,() ) 1, s i ( ) ( ),( u) ( u), w,() v v ) 1 v su u, i ( ) ( ),( ) ( ) ( ),( u) ( u),( ) = X [(1 w,() (1 w,() v v )) 1 s 1 v su u (4) (5) ( ),( u) ( u),( ) ( u) w,() v ( v (1 w v,() ) 1, v su u i, ( u) ( u),( n) ( n),( ) w v,() vs s )] n= 1 s sn n u, And (6) (7),( ) ( ),( u) ( u) Ν = Ζ,() Ζ,() v Ζ v 1 s 1r Su i u After coherent combining using given by,() = + + X =, the fina output is (8) =. Again, assuming 1 to be uncorreated, the variance of,, the variance of,,respectivey can be obtained, as i = (9) i i i,( ),( u) u,( ) = 1,() 1,() 1 p s u 1 v s v v = u p u (,, u ( 1 v ) u u,(),() 1 v s v v u u i, ( u) ( u),( n) ( n),( ) v,() 1 s s vs s u u u, And (0) i n = is given by i,( ), = 1+,(),() i (1) ( j),( u) i,( ), ( ) +,() r + r,() 1r Su u i i, u ( ),( ), ( ) +,() r,() r Si i 4 nternationa Journa of Computer Science Technoogy

6 ,( u) i,( ), ( ) +,() r + r,() 1r Su c= 1 v Sc u i c iu, iu,,( c) i,( ), ( ) ( v),( c) r + v,() η rv i, c,( u),( ) ( ), ( ) ( v),,( ) ( ), ( ) + (,()) Re r + r,() η rv r,() 1r Su v Si 1 s = u i i, u i, c () The average SR on the th subcarrier of the ith user at the output of the rd stage, F SR,() = + SR,(), can then be obtained as () V. SR Anaysis The channe mode used is a one sampe spaced, two-ray, euagain Rayeigh fading mode. Perfect nowedge of is assumed. The SRs in the simuations are measured as foows. For a given reaization of the channe coefficients i, the tota power in the received signa (i.e., power in the LS in Euations (17) (8) is computed, using the nowedge of s s, the desired signa power is computed (i.e., power in the 1st term on RS in Euations (17) (8), the difference between powers in i) ii) gives the interference pus noise power, SR is computed as the ratio of the powers in ii) iii). The average SR is obtained over severa reaizations of the channe coefficients. The difference between the simuated SR the anaytica SR is that in the anaysis to derive the ( ) interference variance it is assumed that 1 i,,,() = ( β1+ ββ) ( β β + β ) 1 4,,() (4). Optimum weight of third stage, in cosed-form Simiary, by differentiating the average SR of the third stage with respect to,() euating to zero, the expression for,,() the imum weights, in cosed-form is obtained, as, ( γ1 + γγ),() = ( γ1γ + γ4) (5) V. Simuated Resut Anaysis The foowing Fig. shows the Bit Error Rate (BER) comparison between the conventiona inear parae interference canceation the proposed weighted inear parae interference canceation at the second stage. Fig. 4. Performance comparison (stage ) The foowing Fig. shows the Bit Error Rate (BER) comparison between the conventiona inear parae interference canceation the proposed weighted inear parae interference canceation at the third stage. 1 Si1, Si, i1 i, are uncorreated, whereas this assumption is not there in the simuations. The average SR expressions in () () can be maximized by differentiating euating to zero to obtain imum weights. A. Cosed-Form derivation of imum weights The average SR expressions for the second third stage outputs can be maximized to obtain imum weights for scaing the interference estimate at the second third stages. 1. Optimum weight of second stage,,,(),,() in cosed- form An expression for the imum weights can be obtained by differentiating the second stage SR euation with respect to,() euating to zero. Accordingy, we obtain the expression for,,() as Fig. 5. Performance comparison (stage ) The foowing Fig. shows the Bit Error Rate (BER) comparison between the conventiona inear parae interference canceation the proposed weighted inear parae interference canceation at the second stage the third stages. nternationa Journa of Computer Science Technoogy 4

7 the performance of the proposed WLPC scheme. The proposed WLPC scheme is shown to effectivey cance the MU caused by the other user CFOs in upin OFDMA at the same time, WLPC scheme (where the imized weights derived are used) performs significanty better than CLPC at the second third stages. t is found that the bit error rate performance of the proposed weighted inear parae interference canceation is amost ten times better than that of the bit error performance of the conventiona inear parae interference at the fina stage. Fig. 6. Overa performance comparison The system parameters are summarized in Tabe 1. Tabe 1: List of system parameters S. System Parameters Vaues O 1 umber of subcarriers () umber of users () 4 Difference between i th th users CFO vaues [-0.1, 0., 0.5, -0.15] [ 1,,, ] 4 4 SR in db 5 db Tabe : Eb/o Vs BER for Stage S. BER for Stage o E b / o in db MF Detector CLPC STAGE WLPC STAGE Tabe : E b / o Vs BER for Stage S. BER for Stage o E b / o in db MF Detector CLPC STAGE WLPC STAGE From Tabe Tabe,We mae interesting observations on 44 nternationa Journa of Computer Science Technoogy V. Concusion The proposed scheme performed CFO compensation in timedomain, foowed by DFT operations mutistage inear parae interference canceation on the DFT outputs. Estimates of the MU for canceation were obtained using soft vaues of the outputs from the previous stages. The MU estimates were scaed by weights before canceation. Expressions for the average SR at the output of the nd rd stages of the proposed scheme were derived. Whie these SR expressions uantified the improvement in output SR from one stage to the next, they were aso used to obtain the imum weights in-cosed form. The proposed scheme was shown to effectivey cance the MU caused by the other user CFOs a comparison between our proposed scheme, WLPC the conventiona system, CLPC is aso shown. References [1]. im,. an, S.-L. im, Joint subcarrier power aocation in upin OFDMA systems, EEE Communication Letter., vo. 9, no. 6, pp , June 005. [] Z. R. Cao, U. Turei,.-D. ao, Deterministic mutiuser carrier freuency offset estimation for intereaved OFDMA upin, EEE Transaction on Communication, vo. 5, no. 9, pp , Sep []. Wang B. Chen, Asymptotic distributions pea power anaysis for upin OFDMA, in Proc. EEE CASSP, May 004, pp. iv [4] M.O.Pun, J. Juo M. Morei, Joint synchronization channe estimation in upin OFDMA systems, in Proc. EEE CASSP, Mar. 005, vo., pp. iii/857-iii/860. [5] M. Toneo,. Laurenti, S. Pupoin, Anaysis of the upin of an asynchronous mutiuser DMT OFDMA system impaired by time offsets, freuency offsets, mutipath fading, in Proc. EEE VTC (Fa), Oct. 000, vo., pp [6] J. Choi, C. Lee,. W. Jung,.. Lee, Carrier freuency offset compensation for upin of OFDM-FDMA systems, EEE Communication Letter., vo. 4, no. 1, pp , Dec [7] Z. Cao, U. Tureh,. D. ao, Anaysis of two receiver schemes for intereaved OFDMA upin signas, in Proc. 6th Asiomar Conf. Signas, System Computer., ov. 00, vo., pp [8] R. Fantacci, D. Marabissi, S. Papini, Mutiuser interference canceation receivers for OFDMA upin communications with carrier freuency offset, in Proc. EEE GLOBECOM, ov.-dec. 004, pp [9] D. uang. B. Letaief, An interference canceation scheme for carrier freuency offsets correction in OFDMA systems, EEE Trans. Commun., vo. 5, no. 7, pp , Juy 005. [10] T. Poet, M. V. Bade, M. Moenecaey, BER sensitivity of OFDM systems to carrier freuency offset Weiner

8 phase noise, EEE Trans. Commun., vo. 4, pp , Feb./Mar./Apr [11] L. Rugini, P. Banei, S. Cacopardi, Probabiity of error of OFDM systems with carrier offset in freuency-seective fading channes, in Proc. EEE GLOBECOM, ov.-dec. 004, pp [1] D. Divsaar, M.. Simon, D. Raphaei, mproved parae interference canceation for CDMA, EEE Transaction on Communication, vo. 46, no., pp , Feb [1] D. R. Brown, M. Motani, V. Veeravai,. V. Poor, C. R. Johnson, Jr., On the performance of inear parae interference canceation, EEE Trans. nform. Theory, vo. 47, no. 5, pp , Juy 001. [14] D. Guo, L.. Rasmussen, S. Sun, T. J. Lim, A matrix-agebraic approach to inear parae interference canceation in CDMA, EEE Transaction on Communication., vo. 48, no. 1, pp , Jan B.Sathish umar currenty is a Senior Lecturer, Sri Ramarishna Engineering Coege, Coimbatore, Tami adu, ndia. e received the Masters Degree from Anna University Chennai, in the year 005 currenty pursuing Ph.D. in Anna University Coimbatore. is research interest incudes Wireess Communication, etworing, Signa Processing, Mobie Communication Muticarrier Communication..R.Shanar umar currenty is a Professor, Sri Ramarishna Engineering Coege, Coimbatore, Tami adu, ndia. e received the Masters Degree from Madras University, in the year 000 the Ph.D. from ndian nstitute of Science, Bangaore, in the year 004. is research interest incudes future broad b wireess communication, Muticarrier Communication systems, Digita Communication, Advanced Signa Processing for communication. e has pubished more than 0 Journa papers in the fied of CDMA systems. is research wor was supported by Swarnajayanti Feowship, Department of Science Technoogy (DST), Government of ndia. nternationa Journa of Computer Science Technoogy 45