Downlink Strategies for MISO TDD MC-CDMA Systems

Transcription

1 Dowli Strategies for MISO DD MC-CDMA Sstems Adão Silva Det Electróica & elec / Istituto de elecomuicações, Aveiro-Portugal Poe: , Fax: asilva@avitt, Atílio Gameiro Det Electróica & elec / Istituto de elecomuicações, Aveiro-Portugal Poe: , Fax: amg@detuat, Abstract is aer roose two advaced sace-freuec scemes wit a re-filterig teciue for dowli time divisio dulex (DD) MC-CDMA sstem We cosider te use of atea arras at te base statio (BS) ad a sigle atea at te mobile termial (M) We derive a multi-user re-filterig teciue tat allows to format te trasmitted sigals so tat te multile access iterferece at mobile termials is comletel removed ad elimiatig te cael dee fades, witout eacig te oise ower ad eeig te M at low comlexit is re-filterig teciue is combied wit te MC-CDMA ad x SFBC MC- CDMA sstems Simulatios results are carried out i scearios wit sectral efficiec eual to bs/z ad usig te UMS cael turbo code, to demostrate te effectiveess of te roosed advaced scemes I INRODUCION As demad for wireless services icreases, more caacit is eeded, ad sice te sectrum is a ver scarce resource, te develomet of efficiet teciues regardig te usage of tis resource are madator e beod G broadbad comoet of wireless sstem must be able to offer bit rates more ta 00Mbs i idoor ad icocell MC-CDMA is oe te most romisig multile access sceme, eseciall i dowli, for acievig suc ig data rates, i order to meet te ualit of service reuiremets of te future multimedia alicatio [] is sceme combies efficietl Ortogoal Freuec Divisio Multilex () ad CDMA erefore, MC- CDMA beefits from caracteristics suc as ig sectral efficiec ad robustess agaist multi-at roagatio, wile CDMA allows a flexible multile access wit good iterferece roerties for cellular eviromets [] owever, te user caacit of MC-CDMA sstem is essetiall limited b te Multile Access Iterferece (MAI) MC-CDMA is for examle studied witi te Euroea IS- 4MORE roect [] It is cosesual tat rovisio of te broadbad wireless comoet will robabl rel o te use of multile ateas at trasmitter/receiver side For te D usig multile ateas at te BS is more feasible ta at te M us, trasmit diversit scemes relig o multile trasmit ateas are ver attractive Sace-time codig scemes, suc as sace-time bloc codig (SBC), relig o multile atea at trasmitter side ad aroriate sigal rocessig at te receiver were roosed [4][5] SBC is a simle diversit sceme wic ca imrove te erformace, ad for te articular case of two trasmit ateas tis ca be acieved witout a rate loss A MC-CDMA sstem ca be desiged i a wa tat two or more adacet arrowbad subcaels are affected b earl te same cael coefficiets ad te combied smbols ca be set o differet subcarriers Ulie te SBC, i tis case te cis are ecoded i sace ad freuec domais ad tis diversit sceme is ow b sace-freuec bloc codig (SFBC) I [6] a efficiet realizatio of SFBC for sstem was itroduced ad its combiatio wit sigle user re-eualizers teciues i U of MC-CDMA sstems Cosiderig DD, aoter solutio cosists i erformig re-filterig at trasmitter side usig te DD cael recirocit betwee alterative uli ad dowli trasmissio eriods [7] Normall tis ricile is valid for idoor or edestria eviromets, ie, i low mobilit scearios e aim of tis solutio is to allow te use of simle low-cost, low-cosumig M Ulie te BS, were comlexit reuiremets are ot so tigt, low comlexit is reuired at Ms, tus ol simle ad sigle user eualizers teciues ca be imlemeted, limitig te MAI cacellatio caabilit Motivated b tis costrait at te M, tis aer rooses two advaced sace-freuec scemes wit a re-filterig teciue for te dowli of MC- CDMA sstems For te first sceme a sace-freuec refilterig algoritm is derived ad for secod oe tis algoritm is combied wit x SFBC MC-CDMA sstem e multiuser re-filterig teciue allows formattig te trasmitted sigal so tat te MAI at mobile is comletel removed ad elimiatig te cael dee fades, witout eacig te oise ower, eeig M at ver low comlexit e refilterig rocessig is desiged usig as criterio te miimizatio of te trasmitted ower at te BS e aer is orgaized as follows: I sectio II we reset te roosed advaced dowli MC-CDMA scemes I sectio III, we aalticall derive te multi-user re-filterig algoritm wic we call multi-user costraied zero forcig (MCZF) for first sceme ad J-MCZF for secod oe I sectio IV, we reset some simulatio results obtaied wit te roosed advace scemes for a sectral efficiec of bs/z i order to assess te re-filterig algoritm i a ig data rate cotext We also comare te roose scemes agaist covetioal x SFBC scemes usig sigle user eualizer teciues suc as, Zero Forcig Combiig (ZFC) ad Miimum Mea Suare Error Combiig (MMSEC) Fiall te mai coclusios are oited out i sectio V

2 rasmitter At User Ecodig Pucturig Iterleavig Maig 8-PS S/P d d Sreader M M CSI Pre-Filterig W M M Mobile ermial At Ci Iterleaver Adder Users At M Users,,- Mod Mod At M At,,m User Decodig Deucturig Deiterleavig Soft Dema 8-PS dˆ Desreader Ci De iterleaver Demod - Figure : Dowli Sace-Freuec Pre-Filterig MC-CDMA Sceme a) II SYSEM MODES I tis sectio we reset two differet dowli scemes: MC-CDMA wit a sace-freuec re-filterig algoritm ad X SFBC MC-CDMA combied wit te same re-filterig teciue A MC-CDMA wit a Sace-Freuec Pre-Filterig Algoritm Figure sows te first roosed dowli MC-CDMA trasmitter for user ad te receiver of te t M After cael ecodig, ucturig ad iterleavig oeratios, te bit streams are maed to a 8-PS costellatio Eac user trasmits P=N c /, (were N c is te umber of carriers ad te legt of te sreadig code) data smbols er smbol e te data smbols are sread ito cis usig te ortogoal Wals-adamard code set We deote te code c = c,,, c,,, c,, were c, is te t ci ad () deotes te trasose oerator We roose to erform re-filterig oitl i sace ad freuec before modulatio erefore, te cis of te data smbols are coied M times i order to obtai M versios of te origial smbols wic are weigted ad trasmitted over te M atea braces us, eac user s smbol is affected b a secific weigt o eac subcarrier ad atea brac Matematicall, we rereset te sreadig ad coig rocess b a sigle vector cotaiig M times te code vector vector of user as [ ] of te corresodig user, ie c = [ c,, c ] e refilterig weigts of user ad a geeric smbol are also gatered i a vector of legt M deoted b [ w w ] w = w were w is a vector of size,,, M tat cotais te set of coefficiets tat weigt te cis tat go to atea m ese weigts are calculated o te basis of cael state iformatio (CSI) accordig to te criteria reseted i sectio III e te cis are iterleaved i order to trasmit tem o distat ositios i te frame We cosider freuec o-selective Raleig fadig er subcarrier owever, due to te freuec iterleavig oeratio, eac data smbol exerieces ucorrelated freuec comlex cael fadig coefficiets icreasig te freuec diversit gai After tat, te sigals of all users o eac sub-carrier ad atea brac are added to form te muti-user trasmitted sigal Fiall, a guard iterval () is iserted to avoid ISI iterferece,m At te M, te sigle atea imlicitl recombies te sigals trasmitted from te M braces at te BS Cotrar, to te covetioal MC-CDMA receiver were sigle user or multi-user eualizers are erformed, i te roosed sceme we ca reduce tis eualizatio oeratio to a simle desreadig oeratio ereb, we ca avoid cael estimatio ad eualizatio o eac subcarrier at te M is results i a ver low comlex receiver desig us, te decisio variable at te iut of te 8-PS demodulator, is for te desired user ad smbol give b, were [ ] dˆ,,, M () = of size M, were is, m te cael freuec resose betwee atea m ad mobile termial for user ad smbol, (o) meas a elemet wise vector roduct e sigal of () ivolves te tree mai terms: te desired sigal, te MAI caused b te loss of code ortogoalit amog te users, ad te residual oise after desreadig B x SFBC MC-CDMA wit a Pre-filterig Algoritm Figure sows te secod roosed dowli MC-CDMA trasmitter for user ad te receiver of te t M As ca be see u to sreadig oeratio we erform te same oeratios as i te first sceme After tat eac sreaded data smbol is weigted b a vector = [,, g g ] g g,,,,, of size, were g, is te t weigt ci of te t user ad smbol We comute a vector weigt for eac data smbol of size, istead M as i te first sceme After te re-filterig oeratio te sigal for a geeric user is give b, [ s ] = [ s s, s,, s ] s = s () of size N c, were = c ( o w o c ) d Desired Sigal c ( o w oc )d c =, MAI,,,,,,,, N C s is of size ad is give b, [ s,, s s ] d c =,,,,, Noise s = o g ()

3 rasmitter At User Ecodig Pucturig Iterleavig Maig 8-PS S/P d d Sreader CSI Pre-Filterig g s s s P/S Mobile ermial Ci Iterleaver s SFBC At Users,,- Adder Users At Mod Mod At, At, User Decodig Deucturig Deiterleavig dˆ Soft Dema 8-PS Desreader Ci De iterleaver r Eualizatio Sace Freuec Combiig Demod - CSI e te cis of te seuece s are iterleaved i freuec domai to roduce te seuece s I tis ew seuece te cis of eac data smbol are searated b ositios i te smbol e maig sceme of te cis of te seuece s for SFBC wit two trasmit ateas is sow i able able : SFBC maig [6] for trasmit ateas Atea Atea Subcarrier s, - s, Subcarrier s, s, After tat, te sigals of all users o eac sub-carrier ad atea brac are added to form te multi-user trasmitted sigal Fiall, a guard iterval () is iserted to avoid ISI iterferece e received sigals o sub-caels ad after demodulatio ad removal are give b,,, were, reresets te freuec comlex fadig cael for mobile, atea m ad subcael ; is te additive wite Gaussia oise o subcarrier sstems are usuall desiged so tat te subcarrier searatio is sigificatl lower ta te coerece badwidt of te cael, ad terefore, te fadig i two adacet subcarriers ca be cosidered flat ie we ca cosider, to be eual to, = s At te M we roose a ver simle sigle user eualizer, wit te coefficiets give b: z, = = s,,, =, = s,,,, /, (4) = (5) ese eualizatio coefficiets are sligtl differet from te oes used b a EGC eualizer e reaso to use te eualizatio coefficiets of (5) istead of te oes give b s Figure : Dowli x SFBC MC-CDMA wit a re-filterig Sceme b),,, =,, EGC, is due to te fact tat wit te coefficiets of (5) we do ot get itersmbol iterferece i te SFBC decodig rocess cotraril to wat would arise usig EGC coefficiets e sigal for o arbitrar air of adacet sub-carriers ad, after usig te SF combiig sceme ad te eualizatio coefficiets defied i (5), ca be writte b, r r After some matematical maiulatios ad cosiderig tat z, we ca write, z, =, After ci de-iterleavig ad desreadig oeratios, we obtai te decisio variable for a geeric smbol ad M, dˆ = c (f o g o c ) d 4444 were f,, = z f is a vector of size, ad eac elemets give b, =,,,, = z,, z,,,, r, =,,,,, r, =,,,, DesiredSigal c ( f o g o c )d =, ,,,, z s,,, s,, MAI MAI,,, z MAI,,,, z z, z, ,, were g, reresets te freuec comlex fadig cael for a geeric smbol, mobile g, atea m ad sub-cael As i te first sceme te sigal of (8) ivolves te same tree well ow terms e re-filterig algoritm roosed i te ext sectio comletel removes te MAI comoet ad dee fades of () i te first sceme ad of (8) for secod oe, resectivel z,,,, z,, Noise (6) (7) (8) (9)

4 III PRE-FIERING AGORIM e aim of re-filterig at BS is to elimiate te MAI ad dee fades at Ms, wile allowig te use of a ver low comlex receiver at M ere, we derive te re-filterig algoritm used i te above roosed MC-CDMA scemes e algoritm is based o zero forcig criteria ad is desiged i sace-freuec domai for te first sceme ad ol i freuec domai for secod oe i order to remove te MAI term of () ad of (8) at all Ms, resectivel Furtermore, it taes ito accout te trasmitted ower at BS, reaso w we call tis algoritm te multi-user costraied zero-forcig (MCZF) I te secod sceme we combie te x SFBC MC-CDMA wit tis re-filterig algorit i tis case to distiguis from first sceme we call oit MCZF (J-MCZF) e iterferece tat te sigal of a give user roduces at a oter M is from () give b MAI( ) = c ( o w o c ) = v, w (0) wit v = c o o, c ad from (8) give b, MAI ( ) = c ( f o g o c ) = l, g () wit l, = c o f o c e weigt vector for M is te obtaied b costraiig te desired sigal art of its ow decisio variable to a costat wile cacellig its MAI cotributio at all oter mobile termials at same time is leads to te followig set of coditios for first sceme, c c ( o w o c ) ( o w o c ) ad for secod sceme, c c ( f o g o c ) ( f o g o c ) = α = 0 = α = 0 () () ece, to comute te weigts for user we ave to solve a liear sstem of euatios, for bot scemes give b, A ( g or w ) = b (4) were A is a matrix of size xm for first sceme ad of size x for secod oe ad b a vector of size, give b, f v0, l0, M M A = v, or lg, v, l, M M v, l, α 0 b = (5) M 0 As stated above te re-filterig algoritms sould tae ito accout te miimizatio of te trasmitted ower ere, te trasmitted ower must be miimized uder te costraits give b () or () deedig of te sceme is roblem ca be solved usig te metod of agrage multiliers [8] After same matematical maiulatios, we obtai te MCZF ad J-MCZF based re-filterig vectors give b, w g = αa ( A A ) b = αa ψ b (6) = αa ( A A ) b = αa ψ b (7) were ψ = A A is a comlex suare ad ermitia matrix of size x for sceme oe ad a real suare ad ermitia matrix of size x, α is a costat used to ormalize te weigt vectors accordig, w = w w = = (8) g = g g = = (9) esurig tat te re-filterig weigt vectors alwas ave uit orm i order to comare wit sstems witout reeualizatio It ca be see from () tat for secod sceme we ol eed to ow te modulus of te cael freuec at te BS Basicall, i tis sceme we ust erform a reamlitude eualizatio i te freuec domai, wile i te first sceme a ase ad amlitude re-filterig is erformed i sace ad freuec domais IV NUMERICA RESUS o evaluate te erformace of te roosed advaced MC-CDMA scemes wit re-filterig algorit we used a idoor Raleig fadig cael, wose sstem arameters are derived from te Euroea BRAN ierla/ stadardizatio roect [9] We exteded tis time model to a sace-time, assumig tat te distace betwee atea elemets is large eoug, to cosider for eac user M ideedets caels, ie, we assume ideedet fadig rocesses e mai arameters used i te simulatios are reseted i able able : Mai simulatio arameters Number of Carriers 04 Sreadig factor Guard eriod samles/time Number of users otal smbol duratio ( s) 6µs Samlig freuec Frame duratio Modulatio Cael rofile 56 / µs (full load) 80Mz s=5µs 8-PS BRAN A Maximum dela/number of as 90s / 8

5 It is assumed tat te receiver ad trasmitter ave erfect owledge of te cael e trasmitter ower is ormalized to oe i all reseted scemes We comare te two advaced scemes roosed agaist Alamouti s SFBC usig covetioal sigle user eualizers (MMSEC ad ZFC) at te M We reset results for a sectral efficiec of η= bs/z i order to assess te roosed sceme i ig data rate cotext sceario e cael codig sceme is te turbo-code defied for UMS, combied wit a ucturig rocess ad a iterleaver to get a overall codig rate of / Figure sows te erformace results for te average bit error rate (BER) as fuctio of Et/No ie, te trasmitted eerg (assumig a ormalized cael) er bit over te oise sectral desit, wile Figure 4 sows te erformace for te frame error rate (FER) as fuctio of Et/No i order to assess te roosed algoritm i data ad multimedia alicatios cotext Figure sows te erformace of te roosed MCZF ad J-MCZF algoritms agaist x SFBC scemes emloig two ateas ad usig sigle user ZFC ad MMSEC eualizatio As ca be observed from tis figure te erformace of te 4x MCZF ad x J-MCZF outerforms te x SFBC scemes usig MMSE or ZFC We also ca observe tat te erformace of te 4X MCZF ad of te x J-MCZF is similar is meas tat for te first sceme we eed more ateas to obtai te same erformace as te secod oe For a BER=E-5, we get aroximatel 09dB ad 5dB gai for 4x MCZF ad 07dB ad db gai for x J- MCZF, we comared wit x SFBC ad usig MMSEC ad ZFC eualizatio, resectivel owever it sould be oted tat we te MMSEC is used te oise variace must be estimated at M terefore icreasig te comlexit e two roosed advaced scemes outerform te x SFBC wit MMSEC witout eedig to estimate te oise variace at M ad for te first sceme we also do ot eed estimate te freuec resose at M Figure 4 sows te erformace of te roosed advaced scemes agaist x SFBC scemes for te same sceario of te Figure, usig te average frame error rate (FER) as metric I tis case we also ca see tat te erformace of te 4x MCZF ad x J-MCZF outerforms all x SFBC scemes e results reseted i te above figures sow tat te advaced scemes are suitable for data ad multimedia alicatios Average BER x MMSEC x ZFC 0-4 X MMSEC x J-MCZF x MCZF 4x MCZF AWGN 8-PS Et/No (db) Figure : Scemes comariso for η=bs/z, BER Average FER x MMSEC x ZFC X MMSEC x J-MCZF x MCZF 4x MCZF AWGN 8-PS Et/No (db) Figure 4: Scemes comariso for η=bs/z, FER V CONCUSION We roosed two advaced scemes wit a re-filterig algoritm for te dowli of DD MC-CDMA sstems, usig atea arras at BS ad a sigle atea at mobile termial We aalticall derived te roosed multi-user re-filterig algoritms, based o a costraied zero-forcig criterio e erformace was assessed for scearios wit sectral efficiec eual to bs/z cosiderig te use of te turbo codes secified for UMS We comared te erformace of te roosed MCZF ad J-MCZF algoritms agaist x SFBC usig MMSEC ad ZFC sigle user eualizatio e umerical results demostrated tat te roosed scemes outerform te x SFBC sceme usig sigle user MMSEC witout beig ecessar to estimate te oise ower at te M for te J-MCZF ad eiter oise ower ad te cael resose for te MCZF, terefore eeig it at a ver low comlexit ACNOWEDGMEN e wor reseted i tis aer was suorted b te Euroea roect IS MORE ad Portuguese Foudatio for Sciece ad ecolog (FC) troug roect POSI/CPS/4670/00 ad grat to te first autor REFERENCES [] Fazel, S aiser, Multi-Carrier ad Sread Sectrum Sstems Wile, 00 [] Sara, R Prasad, Overview of multicarrier CDMA, IEEE Commuicatios Magazie, vol5, December 997 [] IS 4MORE roect, web site tt://wwwist-4moreorg [4] SM Alamouti, A simle trasmit diversit teciue for wireless commuicatios, IEEE Joural o Selected Areas i Commuicatios, , vol 6, Oct 998 [5] Vaid aro ad A Robert Calderba, Sace-ime bloc codig for wireless commuicatios: erformace results, IEEE Joural o Selected Areas i Comuicatios, Vol 7, No, Marc 999 [6] Stefa aiser, Sace Freuec Bloc Codig i te Uli of Broadbad MC-CDMA Mobile Radio Sstems wit Pre-Eualizatio, IEEE VC Fall 00 [7] B R Vocic, WM Jag, rasmitter Precodig i Scroous Multiuser Commuicatios, IEEE rasact o Commuicatios, vol 46, o 0, 46-55, Oct 998 [8] S ai, Adative Filter eor rd Ed, Pretice all, 996 [9] J Medbo, Cael Models for ierla/ i differet Idoor Scearios, ESI BRAN doc ERI085b, 998

6