Inteleave On Pefomance of SCH OFDMA CDM in Fequency Selective Indoo Envionment Suva Sekha Das,2, Muhammad Imadu Rahman, Fank H.P. Fitzek, Ramjee Pasad Cente fo TeleInFastuktu (CTiF), Aalbog Univesity, Denmak, e-mail: ssd@kom.aau.dk 2 Tata Consultancy Sevices, India. Abstact Pefomance of sub caie hopped othogonal fequency division multiple access code division multiplexing (SCH OFDMA CDM) is analyzed and compaed against its non hopping countepat (OFDMA CDM) in fequency selective indoo envionment. The system model is developed and simulation esults fo ealistic scenaios ae pesented. These povide guidelines fo configuing the system and also indicate the suitability of this scheme fo indoo down link ove othe multi caie spead spectum schemes. The outage thoughput of SCH OFDMA CDM is shown to impove ove non hopping OFDMA CDM scheme in indoo down link fo some configuations. The mean thoughput of the both schemes ae found to be same fo both schemes. Effect of block and inteleaved sub caie assignment and the pefomance of a successive intefeence cancelation scheme wee also veified. I. INTRODUCTION The seach fo a suitable access scheme fo next geneation wieless netwoks has poduced seveal combinations of spead spectum and multi caie techniques. Some impotant ones ae pesented in [], [2], [3], [4], [5], [6], [7], [8], [9]. Futue geneation wieless netwoks ae expected to use optimal access scheme fo each envionment. That is, diffeent access schemes may be used fo indoo, outdoo, uplink down link etc. This is because each combination of spead spectum and multi caie tansmission yields an access scheme which can ovecome some paticula undesied physical channel conditions. The chaacteistics ae dependent on the scenaio. The objective of this wok is to analyze a potential access scheme (namely SCH OFDMA CDM) fo indoo down link envionment. The scheme with desied chaacteistics will be developed. Futhemoe it will be analyzed in indoo channel conditions and compaed against a competing access scheme, namely OFDMA CDM. The outage thoughput and the mean thoughput as expeienced by a use device fo both schemes ae analyzed and compaed though simulation. Results ae pesented fo diffeent channel conditions following COST259 [0]. Based on the simulation esults guidelines fo system implementation ae discussed. The est of the pape is oganized as follows. The system model is developed in Section II. Simulation ae discussed in Section III, and conclusion in Section IV. II. SYSTEM A. Envionment and system development A pimay chaacteistic of indoo channels is that it is quasi static, i.e. channels emain almost constant ove long d V Use d Fig.. code code 2 code Vu + + s u b c a i e h o p p i n g IDFT Tansmitte fo SCH OFDMA CDM duation of time. Indoo wieless channels suffe fom multi path fading effect. Multi caie schemes ae vey effective in such scenaios and have been adopted in wieless local aea netwoks [] and metopolitan aea netwok [2]. To suppot multiple uses, eithe fequency division multiple access o code division multiple access is used in conjunction with multi caie systems. Some impotant contibutions ae [] [9]. OFDMA CDM [6], is supposed to be one of the optimal solutions. It avoids multi use intefeence by using fequency division multiple access while it uses fequency divesity by taking advantage of code division multiplexing by speading one use s data ove the allocated sub caie set. This system will suffe fom low outage thoughput paticulaly in quasi static channel conditions. This is because if a set of sub caies allocated to a use duing one packet (numbe of consecutive OFDM symbols) ae in deep fade, they will emain in deep fade fo a long duation. To impove the situation, a highly efficient dynamic channel allocation scheme fo multiple uses is needed. Such schemes ae vey complex and still might not be optimal. A simple altenative to ovecome the poblem of outage, is to use sub caie hopping ove a lage bandwidth. Simila scheme can be found be in [3], which also pescibes fequency hopping with genealized multi caie spead spectum, but it does not give any pefomance compaison of the fequency hopping spead spectum scheme with othes. The sub caie hopping can follow a vey elementay cyclic sequential step incement. It will be shown hee, that such schemes impove outage pefomance ove the non hopping system by a significant amount unde cetain channel conditions. The tansmitte achitectue fo such a scheme can be epesented as in figue. As is shown in the diagam, data symbols fom a use ae fist seial to paallel conveted. Then each symbol is spead. Then chips of
De-Inteleave Channel Equalization Discete Fouie Tansfom the spead symbol ae added and put on subs caie afte inteleaving. It is followed by the invese discete fouie tansfom (IDFT), as is used in OFDM systems. Finally the output is paallel to seial conveted. This is followed by the usual addition of guad inteval befoe up convesion. Figue 2 shows the coesponding eceive achitectue, whee the evese chonology of events of the tansmitte occu. In addition thee is a channel equalize. A geneic time fequency diagam indicating the sub caie hopping with inteleaved sub caie allocation fo a paticula use is pesented in Figue 3. Use V d d De- code De- code 2 De- code v Fig. 2. B. System model synchonized de-otation Receive fo SCH OFDMA CDM OFDM is the basic system on which the code division multiplexing and sub caie hopping is used. Fo the s th OFDM symbol of u th use device a set of sub caies is allocated, which is indicated by u s, whee u, denotes the numbe of sub caies allocated to the use device. K u,s {n} is a mapping function that maps sequential index fom 0 to u onto all sub caie indices in u s K u,s {n} u s ; n {0,,..., u } () K u,s {n} K u,s {n }, fo n n (2) Let v denote the data symbol index in the ange 0 though V u - whee V u denote the numbe of data symbols tansmitted by use u in one OFDM symbol, this is kept constant duing one tansmission bust. c u,v [n] be the n th chip of v th code of u th use. W u is the code length of use u. The n th chip of Sub caies Pilot Tones One OFDM Packet Time, OFDM symbols Data Sub caies of a use Fig. 3. A geneic sub caie hopping scenaio with inteleaved sub caie allocation u th use, fo s th OFDM symbol can be witten as ξ u, [n] = u 0 d u [sv u + v]c u,v [(n)] (3) The sub caie symbol is X s [k] = ξ u, [K u,s{k}] = ξ u, [n] The system needs to be designed such that the guad inteval in the is lage than the delay spead of the channel, this is to avoid inte OFDM symbol intefeence. Anothe design citeia fo OFDM systems is that each sub caie expeiences a flat fading channel. In such situation the the expession fo a eceived sub caie in baseband assuming ideal synchonization can be witten as [4] Y s [k] = H s [k]x s [k] + N whee N is the noise in one sub caie bandwidth. Thee ae diffeent possible combining schemes [5]. With Z s [k] as the equalize coefficient fo the k th sub caie fo s th OFDM symbol, the estimated subcaie is, u ˆX s [k] = Z s [k]h s [k] d u [sv u + v]c u,v [(Ku,s{k})] 0 + Z s [k]n. (4) The ecoveed v th data symbol of s th OFDM symbol of u th use is given as ˆd u,s [ v] = Z s [k]h s [k] u v=0 d u [sv u + v]c u,v [(K u,s {k})].{c u, v [(K u,s {k})]} + pocessed noise. (5) The above expession can be boken into two pats, the useful data symbol pat and the inte code intefeence pat. The useful pat equals W u ˆd u,s [ v] I = d u [sv u + v] Z s [K u,s {n}]h s [K u,s {n}]. (6) n=0 and the total noise tem, which includes the pocessed noise and intefeence due to othogonality loss of the codes in the fequency selective fading channel is ˆd u,s [ v] II (=χ) = k.c u,v [(K whee η = u Z s [k]h s [k] d u [sv u + v] v v u,s {k})]{c u, v[(k } {{ } β u,s {k})]} +η. (7) Z s [k]n{c u, v [(K u,s {k})]}. (8) Sub caie hopping is involved in the above expessions by the efeence to the OFDM symbol index s. By choosing diffeent u s fo diffeent OFDM symbols, sub caie hopping can be ensued. Assuming BPSK tansmit symbol, given that tansmitted data symbol d u [sv u + v] = E s as
P (eo {Z s [k]h s [k]} k us ) = P (ˆdu,s [ v] I < ˆd ) u,s [ v] II ( ) E s = Q σχw 2 u 2 Z s [k]h s [k] (9) (0) whee Q(x) = 2 efc( x 2 ), whee efc is the complementay eo function [6]. It can be easily extended to QPSK and othe modulations [6]. The pefomance of the system depends on the type of equalization combining combination. Diffeent options ae pesented in [5]. The expession fo σχ 2 and the tem in the summation changes with each type of channel equalization Z s [k] and chip combining technique. The pefomance will also depend on the fequency coelation of the channel. Fo ealistic fequency selective coelated fading channel conditions, it is difficult to assess the pefomance of the system without using eal channel models. Extensive simulations ae done with COST 259 [0] channel paametes to obtain the mean and the 0% outage thoughput fo each use fo the hopping and the non hopping systems. The expession used fo computing the nomalized thoughput is ( Be)/N scu, whee N scu is the numbe of sub caies allocated to one use device and Be is the bit eo ate. A. Envionment III. SIMULATION Down link indoo COST 259 channel models with ms delay spead of 50ns (efeed as 2) and 250ns (efeed as 6) ae used. The channel is assumed static duing one packet. It is assumed independent fo each packet. This is a valid assumption knowing that indoo velocities do not exceed 3kmph and packet length limited to 000 octets fo full loading of OFDM symbols, with andom sub caie allocation fo each packet. This is in contast to dynamic channel allocation with channel side infomation at the base unit. Total numbe of sub caies is set to 64. Walsh Hadamad othogonal speading codes ae used. Numbe of allocated subcaies pe use device tested ae 4,8 and 6. The speading gain is kept equal to the numbe of sub caies. Fully loaded system is consideed. When successive intefeence cancelation (SIC) is used, loading is 75%. Fo chip combining, maximum atio combining (MRC), minimum mean squaed eo (MMSE) equal gain combining (EGC) single use detectos ae used. It is found that MMSE pefoms bette than MRC. Theefoe all esults pesented hee ae fo MMSE eceive. Inteleaved and consecutive sub caie allocation (block) also ae veified. In all cases un coded system is used. It is undestood that, standad fowad eo coection schemes, impove oveall pefomance by a few decibels, but the natue of the cuves emain same. Modulation used is QPSK. B. Results and Discussion ) Channel 2, ms delay spead 50ns: In this channel the 50% coheence bandwidth spans on an aveage up to 3 sub caies fo system using 20MHz bandwidth. Mean thoughput (bits/sec/hz) 5 4 6 8 0 2 4 6 SCH SCH ofdma cdm int 4dB SCH ofdma cdm int 4dB ofdma cdm blk 4dB SCH ofdma cdm blk 4dB Fig. 4. Mean thoughput Vs at diffeent SNRs with single symbol detection, fo channel 2 0% outage thoughput (bits/sec/hz) 5 0.65 0.6 4 6 8 0 2 4 6 SCH SCH ofdma cdm int 4dB SCH ofdma cdm int 4dB ofdma cdm blk 4dB SCH ofdma cdm blk 4dB Fig. 5. 0% outage thoughput Vs at diffeent SNRs with single symbol detection, fo channel 2 Figue 4 shows the mean thoughput of the sub caie hopped and non hopped system. They have same mean fo a given SNR (signal to noise atio) and a given speading gain. It is obseved that with inceasing (moe data symbols as well) the mean thoughput is deceasing. This may be because of inceasing inte code intefeence. Figue 5 shows the 0% outage thoughput fo both scheme unde the two type of sub caie allocation. It can be seen that fo block sub caie assignment at small the impovement with sub caie hopping is about 5%. Figue 6, compaes the mean thoughput of using a successive intefeence cancele (SIC) and one using a single symbol detecto. Both systems wee used with sub caie hopping. 75% loading was consideed fo both. Figue 7 compaes the 0% outage pefomance of using SIC against using a single symbol detecto fo the sub caie hopping scheme with block and inteleaved assignment of sub caies. Recommendations that can be extacted fom these
Mean thoughput (bits/sec/hz).005 95 9 85 8 75 SIC 7 ofdma cdm int 30dB 65 4 6 8 0 2 4 6 Fig. 6. Mean thoughput Vs at diffeent SNRs with successive intefeence cancelation, fo channel 2 0% outage thoughput (bits/sec/hz) 9 8 7 6 SIC 4 ofdma cdm int 30dB 3 4 6 8 0 2 4 6 Fig. 7. 0% outage thoughput Vs at diffeent SNRs with successive intefeence cancelation, fo channel 2 esults ae If inteleaved allocation is used, thee is no need fo sub caie hopping. Non hopping and hopping have same pefomance fo inteleaved. This has about 8% bette pefomance as compaed to block assignment fo lage s. If block assignment is used, at low, sub caie hopping is necessay to impove pefomance. Gain of about 5% is achieved in this case. At high s, hopping is not necessay. SIC impoves pefomance in all cases, it is a matte of complexity that may decide its useability. Fo low s, SIC does not give much impovement and hence SIC is not ecommended. At high s, if inteleaved allocation is used, SIC is suggested fo bette pefomance. But, if block assignment is used, SIC has pefomance only slightly bette than single symbol detecto and hence may not be used. Mean thoughput (bits/sec/hz) 5 4 6 8 0 2 4 6 SCH SCH ofdma cdm int 4dB SCH ofdma cdm int 4dB ofdma cdm blk 4dB SCH ofdma cdm blk 4dB Fig. 8. Mean thoughput Vs at diffeent SNRs with single symbol detection, fo channel 6 0% outage thoughput (bits/sec/hz) 5 0.65 0.6 4 6 8 0 2 4 6 SCH SCH ofdma cdm int 4dB SCH ofdma cdm int 4dB ofdma cdm blk 4dB SCH ofdma cdm blk 4dB Fig. 9. 0% outage thoughput Vs at diffeent SNRs with single symbol detection, fo channel 6 2) Channel model 6, ms delay spead 250ns: In this channel the 50% coheence bandwidth spans on an aveage up to 3 sub caies fo same system as discussed above. Figue8 shows that mean thoughput of both schemes fo the two type of sub caie allocation. It shows that at a given, they have almost same pefomance. This scenaio is almost same as was obseved fo channel 2. Figue 9 shows the 0% outage thoughput fo the same conditions as above. Figue 0 compaes the pefomance of a SIC and non SIC scheme fo sub caie hopping scheme. Figue shows outage pefomance fo the same scenaio as above. The chaacteistics ae vey simila to the above case. Recommendations that can be made fo these conditions ae as follows: At low, block assignment with hopping has same pefomance as inteleaved assignment. The gain with hopping in block assignment is about 5%. Block assignment is sometimes pefeed ove inteleaved
Mean thoughput (bits/sec/hz) 95 9 85 8 75 7 65 6 SIC ofdma cdm int 30dB 5 4 6 8 0 2 4 6 Fig. 0. Mean thoughput Vs at diffeent SNRs with successive intefeence cancelation, fo channel 6 0% outage thoughput (bits/sec/hz) 9 8 7 6 4 SIC 3 ofdma cdm int 30dB 2 4 6 8 0 2 4 6 Fig.. 0% outage thoughput Vs at diffeent SNRs with successive intefeence cancelation, fo channel 6 since block assignment may be useful to educe the effect of multiple access intefeence (MAI) due to inte caie intefeence in case of fequency offsets. With high, sub caie hopping may be avoided, but inteleaved sub caie assignment is suggested. Fo low, SIC is not needed fo sub caie hopped systems. Fo highe s, SIC gives impovement in both mean and outage thoughput. IV. CONCLUSION AND OUTLOOK The goal of the wok was to compae the pefomance of the sub caie hopped OFDMA CDM with its non hopped countepat. It has been found that hopping impoves outage pefomance fo some situations while the mean emains same. The gain with hopping can be about 5%. SIC gives pefomance impovement fo highe while it is not necessay fo small s. Using inteleaved sub caies athe than block assignment of sub caies is shown to povide benefit of about 8% in some situations. The esults in this wok can seve as a guideline fo choosing the combination of sub caie assignment, hopping and use of SIC fo indoo scenaios. V. ACKNOWLEDGEMENT The authos ae thankful to Aalbog Univesity WINGlab membes fo delightful discussions on elated topics. The authos ae also gateful to TCS fo funding the eseach wok. REFERENCES [] N. Yee, J.P. Linnaz, G. Fettweis, Multi-Caie CDMA in indoo wieless adio netwoks, in IEEE intenational symposium on Pesonal,Indoo and Mobile Radio Communications (PIMRC 93), Septembe 993, pp. 09 3. [2] A. Chouly, et. al., Othogonal multicaie technique applied to diect sequence spead spectum CDMA systems, in IEEE Telecommunication confeence (GLOBECOM 93), Novembe-Decembe 993, pp. 723 728. [3] Elkashlan, M.; Leung, C., Pefomance of fequency-hopping multicaie CDMA in Rayleigh fading, in Vehicula Technology Confeence, vol., IEEE. IEEE, Septembe 2002, pp. 34 345. [4] Lestable, T.; Husson, L.; Antoine, J.; Wautie, A, Pefomance of adaptive MC-CDMA in HipeLAN/2++, Vehicula Technology Confeence, 2002. Poceedings, vol. 2, pp. 24 28, Septembe 2002. [5] K.Fazel, Pefomance of CDMA/OFDM fo mobile communication system, in IEEE intenational confeence on Univesal Pesonal Communications (ICUPC 93), Octobe 993, pp. 975 979. [6] Kaise, S.; Fazel, K.;, A flexible spead-spectum multi-caie multipleaccess system fo multi-media applications, in The 8th IEEE Intenational Symposium on Pesonal, Indoo and Mobile Radio Communications, vol., Septembe 997, pp. 00 04. [7] QingXin Chen, Elvano S. Sousa, S Pasupathy, Multicaie CDMA with Adaptive Fequency Hopping fo Mobile Radio Systems, IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATION, vol. 4, no. 9, pp. 852 858, DECEMBER 996. [8] Lie-Liang Yang; Lajos Hanzo, Pefomance of Genealized Multicaie DS-CDMA Ove Nakagami-m Fading Channels, IEEE TRANSAC- TIONS ON COMMUNICATIONS, vol. 50, no. 6, pp. 956 966, June 2002. [9] Seung Ho Kim, Sang Wu Kim, Fequency-Hopped Multiple-Access Communications with Multicaie OnOff Keying in Rayleigh Fading Channels, IEEE TRANSACTIONS ON COMMUNICATIONS, vol. 48, no. 0, pp. 692 70, Octobe 2000. [0] J. Medbo, H. Andesson, P. Schamm, H. Asplund and J.-E. Beg, Channel models fo HIPERLAN/2 in diffeent indoo scenaios, EURO-COST, Repot, 998. [] IEEE Std 802.g.-2003, Pat : Wieless LAN Medium Access Contol (MAC) and Physical Laye (PHY) specifications Amendment 4: Futhe Highe Data Rate Extension in the 2.4 GHz Band, IEEE, Tech. Rep., June 2003. [2] IEEE Std 802.6a-2003, Pat 6: Ai Inteface fo Fixed Boadband Wieless Access SystemsAmendment 2: Medium Access Contol Modifications and Additional Physical Laye Specifications fo 2 GHz, IEEE, Tech. Rep., 2003. [3] Shengli Zhou, G.B. Giannakis, Fequency hopped genealized MC- CDMA fo multipath and intefeence suppession, in MILCOM 2000. IEEE, 2000. [4] Klaus Witisal, OFDM Ai Inteface Design fo Multimedia Communications, Ph.D. dissetation, Delft Univesity of Technology, The Nethelands, Apil 2002. [5] R Pasad; S Haa, An Oveview of Multi-Caie CDMA, in 4th IEEE Intenational Symposium on Spad Spectum Tehcniques and Applications, ISSSTA96, Septembe 996,, p. 074. [6] J.G.Poakis, Digital Communications, 3d ed. Mc Gaw Hill, 995.