You are on page 1of 8
Time Division Multiple Access Methods for Wireless Personal Communications TDMA is a classic approach to multiple access in di ital cellular wireless communications systems and is the multiple access technique of choice for several digital cellular and PCS systems. David D. Falconer, Fumiyuki Adachi, and Bjérn Gudmundson feria (Sues ent Compaer so ewer ne naa ore Ra Acs nl Sa Peeing Reach deerme igialcellularandmicrocllular radio stems must incorporate multiple acesschemesthat make efientise Of the allocated bandwidth and radi cellinfrasructre with min mum costand maximum performance [1]. Time Division Multiple Access (DMA) i the multiple access technigue of ehoize for sever al digital cellular and PCS systems: For example, itis currently used in the European, North Amer. ican and Japanese second generation digital cel lular stems (GSM, 1584, and PDC, respectively) 3) and in several wireless personal commun s systems: the European digital cordless system (DECT [2), the Japanese personal wire lesssystem(PHS[4) andthe wireiessaccesssystem proposed by Belicore (WACS [5] or PACS). TDMA as a Multiple Access Technique gestae io stem, a number of wars ina celleommunicatewiththatcell'sbace station which intumcommuniateswiththe resto the work, ge crallyhrough copper or fiber lines. Usually. hepa. ‘cular cellauser terminal belongs tos determined by which base station provides the highest signal power or signal o interference ratio. Ina TDMA, Eelllarradioystom,severaluserstime-share acon ‘mon carter fequen tocommuniate wth the base sation. Each user, transmitinglow bitrate digitized speechor other digitaldat,talloeatedone or more timeslots within a frame in the dewnstream (base tousers) and upstzcam (users to base) directions, asillustrated in Fig. I Inthe downstream direc: tion, the base station brosdeasts to the active tsersin a Time Division Multiplex (TDM) for ‘mat. Inthe upstream direction, each active user ferminal transmits tothe base station only in is ‘own assigned timeslot or slots, Inter-user inter ference is prevented by strict adherence to times lotschedles,andby puardtimesandtine-aignment procedures between upstream timeslots in order {o prevent overlaps due to differen propagation Notethateach individual termina'steesiverand transmitter operates with» duty cycle of VN if there are N user terminals with equal bit rates sharing a common bit stream, Upstream and Aownstream traffic i separated either by using titferent carrer fequensies ie. Frequency Dit Sion Duplex (FD), or by altemating in time, Le. Time Dwvision Duplex (TDD), FDD requires les transmission bandwidth per radio and also less precise synchronization of upstream and down- Seam transmissions to minimize interference (On the other hand, TDD requires simpler radio luplexequipmentand citatesflexblebandwith allocation between upstream and downstream tate. The GSM. IS, PDC, and PACS stems alluse FD, while DECT and PHS use TDD. TDMA js usually combined with Frequency Division Multiple Access (DMA), as diferent carrier frequencies are used in different cele, Frequencies are only reused in cells sufcienty dlistant in order to minimize interference. Fur- thermore, there maybe several carr frequencies ‘sed in cel each with ts own TDMA bit stream And set of user terminals Each timeslot ina frame generally contains wer data its (which may inlade party bts for ero" onttol) as well as extra bits fr synchronization, adaptation control, guardtime, ete Thesmalerthe fraction ofthe irame devoted to these “over head” bits, the more efficient isthe TDMA frame design Forexample,the GSM, PHS and DECT == tems have roughly 30 percent oftheir total bit rates used for overhead, while the IS-54 und PDC ‘ystems have about 20 percent overhead. The ‘major portion of the overhead in the GSM and 15-56 ystems is used for adaptive equalizer tai ingsequences while in DECT, mostothe overhead is used for stem control Fora given amount of ‘overhead, higher efficiencies could in principle be ichieved by increasing the timeslot duration. loweverthscan have the adverse ffetsofineteas- ing the total transmission dela fr time-sensitive traffic such as speech, andior of hampering the ‘stem sabi toadapttorapidchanges nthe prop- "gationenvironment- Reference 2] provides ncomm- prehensive description of the rame structures of Dios. osonss0L00 1995 © IEEE TEEE Communications Magarine*Janary 195 \ {) FOD system: | = _Upsiream frome = ostream: ‘requency = Downsteom frame a Downstream frequency A use fg Figure ¥ Masration of TDMA frames and timeslots several TDMA ssstems, Table | presents com Patison ofa umber of features of contemporary TDMA systems, Figure 1 illustrates possible frame and imesit structures for generic FDD and TDD systems, Comparison with Other Multiple Access Techniques DMA has several advantages relative tater ‘alive multiple access techniques, suchas FDMA\and Code Division Multiple Access (CDMA) Oncadvanagesthatcommonradio and tnodemeguipnent a agvencaserequcney can Reshued among wets at abu station, Another ‘vantage with respect to FDMA i that rites to.and fom cach indus user terminal canbe easly varied aconding to current user needs, by Allocating more or femer timeslot tothe user This s expecially advantageous for integrated service applications, With respect to CDMA, “TDMA has the advantage of much ess stringent power contol requirement since interse ne ference is controlled by timeslot and feauensy allocation stead of by processing ein resulting from coded bandwith spreading Another impor tantagbantgerclatnetoFDMAsnd CDMA tat the timeslot structure gives time for measure ments of alternative slots, requences, and ports imorder to support mobile assisted or mobile contr han. "TDMA also has some disadvantages with respect FDMA and CDMA. Because vse er mina havea UN duce they have periodically Pobstingpoweremelope.Thispreentsachalenge {o designers of portable RF uns. Frequency and fmestot assignment and management ental eran eta comply in TDMA sytem ich isnt found in CDMA systems. Alo the N times higher bitrate means that TDMA may require equalization against mula, which eels shied with PDMA. "TDMA can also be advantageously combined wih packetiype multe sco schemes aint rated voce deta aplication. involving pertaps several diferent types of trafic with different bit ‘ates andonvofeharacterisies, Oe example = Pack et Reservation Multiple Access (PRMA) [2]. {ynamic sot allocation procedure in which idle timeslots are requested by users on # contention basis. I'auser is suecessful in reserving one oF snore slots in a fame, that slot o slots is reserved Tor that user as fong as required. When aslo is relinquished by a user (atthe end of a speech burst or data packet, itis open for contention andreservationtyherusers Asynchronous fer mode (ATM) can also be incorporated in a TDMA-based wireless system [6]. TDMA ist well-understood access technology, whieh has Deensuceesfully wedina number ofwiredsndwite Tess digital transmission systems. This has n0 doubt contributed tits adoption i second: gen- ration digital cellular and PCS systems. Objectvecomparisonsofspectrumetficieneyor capacity of FDMA, TDMA and CDMA systems fe difficult to make sine its almost impossible to have similar assumptions for different systems. Fusthermore, most comparsonsare madebetween siforent systems, at diferent evlutionary stages, land not between the access methods themselves. Ione were to compare two optimized systems, with diferent access methods, theircapactieswould Tikely be similar. Capacity in a celllar based per sonal wireless communication systems can b ‘expressed as cll eapacity or system capacity. Cell ‘apaciy, orradiocapacity dependson teas ink performance, ie, athowlowenrertoinererence Fatio (Ci) the system can operate at acceptable quality. This in turn depends on the system Fequirements for speech gualty (fora given dig {al speech compression technique) or data reli bility Cellespacy smeasuredin Eang/MH2Ce, ands adiionall used when analyzing the capac: ity ofa sjstem. Both the modulation efficiency (bits per second per Hz) and the frequency reuse factor must be taken into account. Sytem capaci, onthe other hand, depends on the feasibility of Supporting small els (micro cells and pico cells) andthe cell capacity. System capacity measured in Erlang MHikmwhich means quite naturally that the eapacity increases with decreased cel sz TDMA isa well-under- stood access technology, which has been successfully used ina number of wired and wireless TEEE Communications Magazine Janus 195 1 {WFigure 2. Mlusation of Rerarchical cell sructoes, which shows a stem with ive layers of ell one macrocell ose coverage, one sree microcell 0 lncrease capacity anda number of indoor picocell A hee celle give diferent Ips of coverage othe same geographical area ACA alo fguenty called Dyramie Channel ‘Aigumen (DCA) Capacities fr varios types of cellular systems can also be compared roughly and simply,on the bass ofthe numberof ser terminal they can Sup portina fired bandwidth given the Cilrequrements ‘consider a com ‘stem with the North American AMPS analog cellule sys tem. Theltter aecommodatesoneurerina30-KH2 bandwidth requiresabout ISdB.CTforsatisfacory quality, and gencrally requires acellular frequen ‘Sjreusefactorof with bacctoranfennas The CISM ‘Sistem, aided by slow frequency hopping and coding. requires about 9 4B Ci, a frequency fuse factor of 3, wth 3sector antennas, and in Ii inital version, accommodates eight users ina 200-KHebandwadth,Tasthe GSM capacity sabout (0x7) divided by (3x 20008) = 28 times that of ‘AMPS, Similar calculations [7] suggest thatthe {5-54 digital celular system ean exceed the AMPS capacity bya factor of about 38 0.63, ad that the PDC digital cellular system can exceed the [AMPS capacity by a factor of about 4.21076 Each of these systems factors can (and will be approximately doubledty the use othalfratespeech coders, These upper andlower bounds on improve ‘ment factor correspond topessimisicandopumistic Frequency ree laciors of? and 4, respectively, Capacity Enhancement and Evolutionary Scenarios Betiasicts(oarocesoruniretets)and mallcellsae neededsimultancousy Te age cells are used to get coverage, and sometimes Iso to upport hand-offs between small cells, ‘The small cells are used to achiewe high capeciy, to get coverage in certain places, eg. tunnels and 65, and also t provide accesso users with {ow power portables This whole scenario can be deseibed as hierarchical cell sutures (HS); as ‘Shown in Fig. 2. Th definition of HS i that ee {erent cal ypes exists simultaneously i different call ayers, covering 2 common geographical area. ‘To get very high system eapacty the requirement is that HCS mast be supported [8] Before discussing the consequences of HCS for TDMA systems, se look ito the possibilities ‘of increasing the ell eapacty via either or both ‘of Adaptive Channel Allocation (ACA)! and Fre ‘quency Hopping (FH), The idea of ACA Sto avoid imterfeence by allocating channels based onthe instantaneous signal to-interferene station. This ‘means tha there no fined euse distance. Some” times theco-channeluserisveryclose;eg.whenthe Imabile portable is close to the base station Sometimes the operating conditions for one aser sreveryificule then the co-channel sere are forced tobe distant Also when the traffic conditions are varying, channels can be “moved” to the peak trffie cells The capacity increase with ACA can be as much a5 100 percent, compared tothe same system using a fixed channel allocation. [a [9}, [ACA\te shown tobe able to increase the capacity of 1S-54 on the order of 30 percent. ACA is also part of the DECT standard (10) “Another way to increase the cell apaciy ito use random FH, which ean either be fast, (hop. Ping rate on the order ofthe bit ate), or slow, (hopping between busts of data). Slow FH (SFH} ‘spar ofthe GSM standard. Ina SFH system, the ‘ser inside ace have the same hopping sequeace, ‘ut offset from one another such that they donot interfere je, he users within any cell ae orhog ‘al, The hopping sequences in diferent cell are random, which means thatthe interference from ‘ther cell users isaverage. The averaging is made possible through iterleavingsnd powerfuichannel Coding InGSM forerampe.theintereavingsdone ‘over Sbursts, and a convolutional code of rate R 1/2. and constraint length K = 5,8 used [11] The combination of interleaving and channel cralng give the effect that a number of bursts can be completely wiped out, and the information ean still be recovered. In OSM, approximately «wo ‘ut of eight bursts can be wiped out without oss ‘of information. See [12-1] for further informa- tion on capacity in FH-TDMA systems "Note that a FH-TDMA system relly can be considered as a CDMA system, withthe code 35 the hopping Sequence. The increase in capacity ‘somes fom the fact thatthe interference is ver Sed, which means that one ean design the 3 {em closer othe average interference situation, ‘compared to non-hopping systems. The frequency Teuse factor can be reduced to one, by reducing the ell load (fractional loading). This very sim Hae to direc sequence (DS) CDMA systems, with theexabenettthattheusersinsdeacellareorhos- ‘onal to one another. ‘Another benefit of FH is increased robustness (alversity) against fas fading, dv to multipath propagation, expecially for slowly moving users ‘The principle thatthe total hopping bandwidth is sufficiently large that the different hopping frequenciesarenottoohighlycorelaed. Improved performance results because only 3 few of the Frequencies wil be subject to deep fading at the “Asdiseussed earlier, coreallyinereaethesystem 2 TEE Cormmuniations Maga «January 195 capacity. a hierarchical cel structure (HCS) is {eguired und ot only for capaci, but aso for Corerage and operational aby To sappoc CS, ‘ele handfbetweencellayersisteued Fue thermore. hand-ff shoul no be forced between Cell layers should Be up 1 the operator to ‘howe the hand-off strategy [8]. In HCS systems ihe ferent lyers need to ue diferent RFs ers so te diferent cell ypes are orthogonal ach other-This is due to Rea-far effets ep tising tow power miro cells and high power macrocelisvand the requiement that hand-off Should not be forced between layers. Mobile {sisted handoff (MAHO) is therefore requited between different RE-cariers. This the normal operating made for TOMA systems, as shown in Fig. here measurements om oher RF-carcers te made at the mobile station during the fee time after the receive and transmit Dutss. As indicated in Fig. 3. the mobile measutes one «oe frequency ech frame, cing hebugh al possible aerate requenies during he suces on of frames. The MAHO proces ar ofall ‘TDMA standards The evolution of TDMA systems not only meansincreasedcapacg, Probablyevenmorimpor {antisthe aspect ol nctease Tenby nfrequeney planning. The use of ACA means that the system Seeltplonning continually sapting to chanelng conditions. Modulation and Detection Approaches Coarse egueneysese factor ‘nd Ci (carer ointeterence ratio) required bby multilevel modulation schemes lead tothe Conclusion that four-level modulation schemes Such as OPSK of its variants ate appropriate ‘hoices in many cellular environments] Beyond four levels in hese ervironments, the effect of ‘dereaved required bandwidths more than offset, bythe larger frequency reuse factor needed for stfiient © Homever te reasoning tn [1] leads Frequency oan) .- two | a 3) on) | . \ ? TOMA tame a Figure 3. MAHO process (RX-TX measure): (TX), (RX) and tM) are the Time monsmision respon and measurement frsguencies on base station 9) respectively nthe moe aon to the conclusion that modulation schemes with more levels ma be more appropriate if eels are physcaly isolated by walls or ithe stenuation ‘eus-distance exponent sigh sivor more). Mul level modulation also usefulin proposed systems ‘whose instantaneous bit rate can vary to accom tmodate fading. An etample s differentially coded amplitude phase shit keyed modulation (DAPSK or sar QAM) [15 which can e detected ‘coherently or diferent “Table { summatizes several representative current TDMA second generation cellular stems Including theie modulation parameters. Most use ‘wo-orfour level modulation schemes. ASexaMPES ff four-state modulation applications. the North ‘American [16] and Japanese digital cellular ss tems [3] use x/4-shifted differential QPSK (DOPSK), with either coherent or differential detection The GSM and DECT systems use ‘Gaussian Minimum Shitt Keying (GMSK), a hich fan be considered a form af binary frequency Ht Biota 2708 x Jenni laze 152 is] 384 Bondwisth eae space) —_|200 ie Bo KH a5 wie 172 whe 3008 Tie ot daton 0577 ms 7m evs ost? ms [085m Unsreom sper fore [8 16 with arate code [3 (6 wih halfete ode 3 (Swit haatecoded) [12 4 Speech coxing ss eT asian vse (67 Kon vseur fae [ame | fabeem | Aorcat 00 9 100 Fo 00 oo jroo | 0 ‘Percent papadin ie sot 173% cry jsow lam [rw | cindex party bs _| ‘Modulation use =a pork na DOK fans | 4 0aesx coaing 3 dase: Cove = 12 [3 clases: Cove = 172, [2 dass Coded <7, [RE only CRC On = comes or coded” |ReSconal: or coded |= come) + CR oF fsenGoruncoded —[+eRG oruncoded —_[uncaded ‘opie equaeer Mandatory Mandatory (atonal ene [none | I Table 1. Chris of comemporan TDMA miter 7 ro . xe we pls behice sano Mier ‘detector xy sas ink eel Figure 4. Opiimal posrdetecrion diversi receiver ‘modulation, or also as {is modulation bandwidth etfciency is loser to that of OPSK thant that of BPSK), The PACS temproposedby Belleore[S.TJusescanerentOPSK. “The choce of coherent or noncoherent mod lation depends on the relative ume variably of the radio channel and on cost and processing power considerations, Condition of higher time ‘Variability tend to favor the use of diferential oF ‘other noncoherent detection. Lower time var. aly and more singeat performance or spect ‘ffcieney requirements fend to favor coherent detection approaches. Furthermore, reference ‘ireted equalizer adaptation algorithms in effect, supply the carier phase estimate. tn this case it ‘equites very ltl extra hardware complexity 10 Include coherent demodulation Differential detection of diferenilly encod ted phase madulated signal sa form of none. herent detection which hs the advantages of fas synchronization, high robustness to multipath fading, snd reduced hardware complexity. Con ventional differential detection for Mary DPSK ‘detects the phase difference between the fo su ‘essivelyrevewed signal samples and then deciles which symbol was transmitted. Since the phase reference iscoruptedby nase andinterftence the biterror rate (BER) performance i inferior {0 that with ideal coherent detecto Decision feedback ditferential detection is simple and practieal method af improving the formance of differential detection to approach ‘hat of coherent detection{ 18} Feeding ack the past L detected symbols, it reverse-modulates the Stored. pas received signal samples and gener ates a reliable reference signal. Ifthe past de sions ar all correct, the reverse-modulated ign ‘components at all in phase and canbe coherentiy added. This modulation and addition process increases the reference signal SNR by 10 og Te fast synchronization property of differen detections sill maintained Differential detection ‘s susceptible Co frequency offset between the "tansmiter and receiver, the frequency offset can be estimated by corzelating the detector out- [put waveform with te feedback detected symbol Sequence. Thenthe detec outputs inversely phate forated before he decision cicuitsoastacancslthe effect ofthe frequency ose, Delay Spread Mitigation Requirements Dots 2228 tie dispersion sw physical ptenomenondueto mulipath propagation The transmitted signal wil travel through 2 multipath emironment and wil aie dispersed nthe eee, The ems delay spread is a measure ofthe standard deviation othe chanaelimpulse response duration, Intends toinereasewiththe propagation path ength and antenna heights. Thus t= typically higher foe angel systems than for indoor miroceliar systems. How much intersymbol interference (IS) ‘given amount of delay spread causes depends on the symbol time. Ifthe ms delay spreat exceeds about 10 percent to 20 persent af the symbol ime. delay spread mitigation is necessary (19, Uypically either diversity or equalization, or both, TDMA radio system with lrger relative delay spreads, where the transmission bandwidth sign icantly exceed the channel’ coherence bandh, can achieve a form of frequency diversity by wing ative equalization. The same king of benefit is realized by argesbandwidth CDMA systems which use RAKE recetvers. The resulting relative ‘dues gainwouldbeexpestedtobehigherfor CDMA, ‘ystems when the channel delay sreadiksmalleom- redo the datasyaboiatrval butmight be rough Iyequalfor TDMA and CDMA systems hoxe sign bandwidth greatly exceed the channel's coher ene bandwith ‘The GSM and IS-58 digital cellular systems both cequte adaptive equalization The require ments on delay spread mitigation are defined uitedifferentyin these systems, InGSManumber of typical channel models (delay profiles) have been postulated. Roughly speaking, in GSM. Impulse responses of any shape, with width of ‘upto around 18-20 us, shouldbe and. Tn TS-54 the different changel models are characterized ss ‘two-ay model, with ray separation «para, ‘The maximum separation between the two Cays Is 40 us: one symbol time ‘The ISi'in the (wo systems wil also differ quite significantly, and consequently the equate Er requitements, Witha time dispersion width of Is usin GSM. 5 bits might interfere with one another. GSM isan example of a TDMA system ‘whose transmission bandwiath typically exceeds the channel coherence bandwidth by a consider. able amount In 8-58 two symbols wil interfere, Since the symbol time is the same asthe time Aispersion width requirement (40 ms), Further spreading results from transmitter and receiver brand limiting filters, bur the ISI span in symbol intervals is still less than that of GSM. So the GSM equalizer iseit will be more complex than that for 15-54 Delay spread conditions requiring ‘equalization andor diversity wll lsa be encoun tered in most indoor wireless systems operating sith bitrates above about 2 Mtv Diversity and Channel Coding Aisszrs ivi tsein i basa cas fed into two types: predetection and postde- tection combining Predetection combining ether egies the careful co-phasngof al the received Signals tha experience fat random phase variations, ‘orisdone simply by selecting the "bes ofthe anten. nautpusfor demodulation and discarding the oth {rs Postdetection diversity ssimpler fo implement than co-phasing (but more complex than selec: tion since it requires the duplication of receivers) and matches noncoherent receivers [20}, since all the differential detector outputs can be easly ‘combined; thsi because the detection process ‘Femoves the random phase variations TEE Communications Magarin *Janutry 195 A block diagram of an optimal postdetection versity ecever withlifferental detection isshown, In Fig. 4. The detector outputs ate weighted before combination according vo the time varying channel condition of each branch, The weight is In proportion to the squate ofthe received signal envelope R(faimiterisnotwsedbelore diferent etection, no weight function is required) to ‘weaken consbutions from weaker signal branches Performance results showninFig-S,weremeasured for independent Rayleigh fading of desired and interfering signals, witha normalized maximum Doppler frequency FT = 100125, heresulsshow that aI percent BER can be schieved at below a sverage SIR of [4 dB (which s about dB small: ercthan with selection combining) with simple ‘wo-branch diversity. Thispermits frequency reuse factor of four ia thre-sectored cel is use. ‘Sinee ISI due to delay spread becomes larger asthe signal becomes weaker, diversity combining fan reduce the effect of delay spread. The Japanese PDC system relies on dversiy to combat ISI, and generally doesnot use adaptive equalization For example, the tolerable rms delay spread for | per= ‘ent BER can be increased to near 20 pergent fof the symbol time by using two-branch diversity. ‘This corresponds to 9.5.8 rms delay spread for the Japanese PDC using 21 ksymbols. To cope ‘with delay spreads beyond thivvalue, diversity ‘an be combined with base station antenna bear tilting weaken multipath signals eavelling from long distances. ‘Convolutionaleodingisapowerfultechnique for able data transmission in mobile fading env ronmeats. Decoding ca be applied tothe ouput ofthe postdtection combine n Fig. 4, Sot cs Son Vitebidesoding wth ntereavingisiself form ‘of postdetection combining, and thus can increase the effective diversity order without increasing the number of antenna elements. This fof prac ticalimportance, especialy forband-held portables Foridealinterleaving,theequvalentorderof diser- sity isthe number of space diversity branches Timestheeffectve cove length (ECL) thicmaybe ‘elinedastheshortestmbolerroreventpath length ‘ofthe eode [21 The theoretically predicted relative cel capac- ity of DPSK with iwo-branch space diversity and constraint length K = 5 punctured codes, ‘compared toa system with no diversity and no oding, is shown in Fig. 6 for BER=10' [21] ference but no noise, hexagonal cell ayout, ideal bit interleaving, zero delay spread and doppler ‘Wilustrates the substantial contribution (0598 tem capacity ftom coding ad diversi. Symbol orbit) intereavingisusedo change slow fading into forced fst, Independent fading veld the maximum power of channel coding. For ideal interleaving the required (the transmission time sep consecutive coded symbols), is on the order of ‘O.38fp., where ps the maximum Doppler tre~ quency, determined by the terminal speed and amir wavelength. Tis corresponds t0 4 {ime and 73 synbos fora transmission cate of 16 symbols as 900 MHz carrier Frequency and 100 kan terminal speed. shorter interleaving depth, about half the above value, can be used Wwithaslghtperformanee degradation Theeffectof EEE Communications Magazine * Janey 1995 Figure 5. BER performance. ‘Average SIR (48) 104 Two-branch Single-branch The GSM and North American digital cellular systems have chosen adaptive equalization, applied at the receiver, to combat severe multipath delay spread. Imperfect interleaving i similar to that of fading correlation ina space diversity system. Fora fre ‘quency selective high it rate indoor wireless channel, coding, even without interleaving, cen also help to reduce the order of divers tolimit outage probability [22] Adaptive Equalizer Techniques Ti 956 and North American itl slr systems have chosen adaptive equalization, applied at the receiver o combat severe mult path delay spread. The wireless channel's fe ‘quency response is often severely distorted by multipath; 9 result, decision feedback equal ization (DFE) ora version of maximum likelinood sequence estimation( MLSE) arewsedinpreference to linear equalization. MLSE equalizers with 32 sates[23] and DFE [24jequalizershave been mpie- ‘mented for GSM systems with delay spreads on the order of Sbitinervals‘Thechoicehetween MLSE. and DFE is sualyaperformance;complexit trade fff. MLSE equalizers are inherently better than DFEs, but are generally more complex. However, felatvely simple versions of MLSEs, which han ‘le truncated impulse responses have been shown tobe very efectne for the North American IS 54 ‘ister [25-27] I should sso be noted that adap- tive equalizers also have an ability to suppress co- channel thesame “The function ofthe equalizer’s adaptation algritnm sto sequireand racktherequredequal- izer parameters. These of parameters that must be estimated for a MLSE equalizers the sam: pled channel impulse response. The DFE requires Ssetofforwardandteedback apcoeficientsuhich an be adapted directly, or computed from the estimated channel impule response Since ina TDMA system the channel response tor from a particular terminal may have changed ‘gnifianty ine the lastimeslo, acquisition of the ‘equalizer parameters must generally start afresh ineachtimesloe Typically the aequsivonalgorthm isbased om reference data symbols derived from ‘training sequence of known data symbol, ‘Adaptation algorithms with very fast convergence fare required due tothe very shor taining per (ods (26b for OSM, 28bforIS 54) Theclasi ple is recursive least squares (RLS). In GSM and [sth taining sequences are designed to have impulseslke autocorrelation fanetions,s0 that 3 simple correlation operation estimates the chan nel impulse respense [23 ‘When it comes to channel tracking an impor tantisueiswhetherone hasto track duringtheburt, ‘ormot. Withacarter frequency ofaroundD Ma, and assuming a vehicle speed of 100 kmh, we get maximum doppler frequency of fp = 83 Hz, ‘The minimum time between two fading dips i= therefore roughly 6 ms (1/2p). In OSM the Burst Jength fs atound 0.38 ms, and furthermore, since thesynchronizationwordiin the mile othburst, the maximum variation ime is around 0.29 ms ‘This means that the channel s practically constant ring the burst. In 18-54 on the other hand, the burst length is 6.7 ms: i.e. ofthe same orders the time between dips. This means that one will, ‘experience severe channel variations during burst 8nd highly optimized tracking algorithms are needed. On the other han, for broadband indoor wireless systems, where dopplerissmallandthe data ‘ateisrlatively large, trockingandcoherent demo ‘lation pose lite problem. Fast tracking algorithms utilizing receiver ‘decisions toprovdeareference introduce problems ‘of error propagation; decision errors ean rapidly Girect the equalizer parameters towards the ‘wrong values. Furthermore, inthe case of MLSE. ‘ualizers.the MLSE decsonalgorthmintroduces 8 Significant delay inthe receiver's decisions, ‘whichcan hamperafast decision directed algorithm. ‘Oneway ocrcumventthese prablemsistodspense ‘with eesion-directedtrackingaltogether estimate the channel impulse response or equalizer coef cients during periodic taining periods, and intr- ola these estimatesbetween tainingperiods, This Strategy has proven effective and bust for suff ‘ently requent raining interval 29) Ieanteused for example inthe base to mobile diretion in IS- 1 phasealignment procedure 0 allow for Conclusions T BMA se asic aporoach o mate sesee in digital cellular wieless communications ‘ystems. Wehave summarized number offrequency nd timeslot allocation techniques for enhancing thecapaity and lexbily of TDMA based systems 8 how the problems of fad ing, delay spread, time variability and interfer. ence afect TDMA systems, and how they may be countered and even exploited by appropriate fechniques of detection, diversity, coding, adap tive equalization and slow FHL. Tt is worth empha sizing thatthe use of one ofthese techniques, Slow random FH, results in a system thats in effect a hybrid of TDMA and CDMA. References cae (a0 est (Siete anette 8 ct he rot ve Wc 4 ha TS "a spt an eS Dg ae (OER hn op con the Rao EEE Communications Magazine * January 195 Itis worth emphasizing that the use of slow random ‘frequency hopping results ina system that is in effect a itn fama ae hybrid of ister TDMA and CoE nina seth tn egsnes CDMA. Upcoming Features in | Wireless Security IEEE Communications Magazine | Engineer February 1995 soon petro oi sarce nf serie onl tostqpr ct atceronmmiaong basses are xn toe comp and | te iS Sati eit roesenspowlerin eS i sence npr rane wih © A Multi-faceted Approach to ‘asf menl opr aid Forecasting Broadband Demand | tehiclgtte atti teased wins ‘Segue tu oT ‘Sei Spuring STEPS CTE Mob and CTA a and Traffic | UD Fras Speak fd deslopment ert 2 wel as age ar afte soiree sesso | “oui, andes mst ave an MS or PD a EE and | Methods of Forecasting Long Term ‘Setupatatens alr eaphaseen daa Seat crmomhynavielesrnen. ina DOS/UNDK enveonmentis Demand for Wideband and | Proficiency in Mio required efter a ents pckag including an | Broadband Services in the ‘We gr an cunning bones pstogs incncing 3s. Please send Residential Market ne fea Sivan Road, Forecasting Broadband Demand | Between Geographic Areas TEE Communications Mapai + Jeusey 1998 3”

You might also like