You are on page 1of 4
A Study on time domain Atsushi MATSUMOTO’, Kenic! '"Matsuthita Communication Kanazawa R&D Labs. Co,Ltd 14 3, Sainen, Kanazawa, Ishikawa, 926-0024, Japan Phone: +81-76-224-1283 FAX: +81-76-224-1243, E-mail: matumoto@krd.mcimei.cojp Abstract This paper presents computer simulation results on spread- ing in both the time and frequency domain for OFDM- CDMA. The results show that the suitable layout on which spreaded chips were mapped was different by some condi- tions (eg, number of multiplexing codes, propagation con- dition, et). For example, in case of employing a number of ‘multiplexing codes, spreading onby in the time domain out- performs other layouts. On the other hand, in case of half ‘multiplexing codes, spreading in both the time and fre- ‘quency domain oupperforms the others. Keywords ‘OFDM, CDMA, Spreading, Time and Frequency Domain INTRODUCTION In order to accommodate the very high data rates that next ‘generation cellular systems will be expected to deliver, @ number of systems have been proposed. One of the pro- posed configurations is OFDM-CDMA. This is based on @ number of narrow carriers that experience non-selective frequency fading on which orthogonally spread symbols are modulated The spreading can take place either in the time or fre quency domain with the latter being the candidate most frequently investigated. In this paper we present results showing the benefits of spreading in the time domain and ‘we also explore the option of using both the time and fre- quency domain. ‘SPREADING ON OFDM Different symbols belonging to the same user are spread using orthogonal codes. Orthogonality depends on all chips comprising a codeword being received with the same power. Chips modulated on carriers sufficiently separated in frequency will be subject to the effect of frequency se- lective fading. This results in the loss of orthogonality and ‘mutual interference between codes. It is much easier to satisfy the constraint thatthe power of | the chips in a symbol do not vary when spreading is done in the time domain. The system under investigation has a symbol duration of 7.3Sus which means that even at speeds of 200km'h a fading cycle comprises 92 symbols. This ‘means that al the chips ina symbol carried on a single car- 0-7803-7442-8102/517.00 © 2002 IEEE 725 spreading for OFCDM MIYOSHI", Mitsuru UESUGI"*, Osamu KATO" **Wireless Solution Laboratories Matsushita Communication Ind. Co,LTD. 5-3 Hikarino-oka, Yokosuka, Kanagawa, Japan Phone: +81-468-40-5439 FAX: +81-468-40-5183 E-mail: Kenichi. Miyoshi@yrp.mei.ei.co,ip rier are likely to be received under the same conditions helping maintain orthogonality The fading for each subcarrier may be constant in the dura tion ofa slot (coding interval), but the fading can be very ifferet for the various subcarriers. So, when spreading in the time domain the use of an interleaver becomes neves- sary in order to disperse burst errors from subcarriers re- ceived in low SNR conditions. It is also possible to utilize both difnensions ~ frequency and time — at the same time so as to optimally take advan- tage of channel and traffic load conditions. PERFORMANCE EVALUATION ‘The configuration used for the link level simulations is shown in Figure 1. Fourteen symbols are spread by ot- thogonal codes with spreading factor of 16 and modulated fon the same subcarrier in the time domain. slot has a time duration of 0.25ms, and contains 34 symbols. The first and last symbols per subcarrier in each slot are pilot sym- bols used to estimated the channel and perform coherent demodulation. Less than 6% ofthe ttal base station power is allocated tothe pilot symbols. Figure 2 shows the different spreading options that were ‘evaluated. Spreading in both time and frequency domain hhave been simulated. The first spreading patter (a) de- scribes the spreading only inthe time domain and the last spreading pattem (e) deseribes the spreading only in the frequency domai ‘The block diagrams ofthe simulator’s transmit and receive chains are shown in Figures 3 and 4 respectively. Table 1 lists the simulation parameters. The channel model consists of 12 paths, each path arriving with 1dB less power than the previous path and at intervals of 115ns. The results show that spreading only in the time domain leads to minimum power requirements for a given packet error rate. This is true both for 14 multiplexed codes with spreading factor 16 ~ Figure $ ~ and for 30 multiplexed codes with spreading factor 32 ~ figure 6 Figure 7 shows the Eb/NO required to guarantee a PER of 1.06.2 for each of the spreading options in case ofthe three ‘ways of multiplexing codes (ie. 1 code, 8 codes, 14 codes). In case of 14 codes, spreading only inthe time domain re quires 1.84B less in terms of received Eb/NO when eom- pared to frequency only spreading. On the other hand, in case of 1 code, spreading only in the fequency domain requires 3.84B less when compared to time only spreading. ‘Additionally, in ease of 8 codes, the layout of §32 (Fig.2 pate () tequires about 1.043 less when compared to frequency or time only spreading. Similarly, these ae tre for I, 16 and 30 multiplexed codes with spreading factor 32~Figure 8~ These simulation results show that spreading only in time domain can reduce the interference between spreading codes compared with frequency domain spreading. This is explained by the fact that the power variation ofthe chip comprising one symbol becomes smaller in the time do- main spreading method compared tothe fequency domain spreading. Conversely, since there is litle diversity gain in time domain in case of low Doppler frequency, the fre quency domain spreading outperforms spreading only in time domain because of frequency diversity gain if muli- plexing codes are not employed. Moreover, since the inter- ference of spreading codes and the effect of frequency di- verity has a trade-off relation, there are suitable layouts — in both time and frequency domain on which spreaded chips are mapped. In other words, system performance improves by changing the spreading layouts at every condi- tion ‘Table 1. Simulation Assumptions Parner Expanation/Assampiion Bada SNe ‘amber of 5 caer 32 Spreading Code OVS Scrambling Code Gal Code Spreading Factor 16,32 ‘Number of Multicode 1,8, 14), (1, 16, 30) path (Channe Move! Exponential model Path interval= 11505 asian Doppler ie a ueney == Turbo Coding. RT R13 So Ome OFDM symbol eng TED Guard nerve eth Tao Power Conf Ta contaled Channel entiation | Eniate with pot Despreading MIMSEC rato ar nanited | 2348 69%) 726 tie Plot SF:sproding Factor frequency Ey Number tM code Figure 1. configuration of only time domain spreading (freq, time) = 1,16 time domain spre (Aroq, time) = 8,2 Figure 2. optional spreading variation (SF=16) CONCLUSION ‘The performance of an OFDM-CDMA multiple access has ‘been investigated with.respect to the dimension on which spreading is performed. We found that due to smaller intra- ‘symbol variance in the time domain, @ higher degree of crthogonality is maintained if each symbol is carried on @ single subcarrier, leading to lower intersymbol interference and better performance. Spreading in both time and fre- quency could offer a reasonable compromise between the interference and the diversity effect. a * > time Figure 3. path model Figure 4. Block Diagram for the link level simulation 18600 — e100 seo sE01 & & 1602 see 1e03 1e03 > 5 8 6 mw o 8 © % 0 2% erage EONOEB) average EDNOAB) Figure 5. PER every optional spreading Figure 6. PER every optional spreading (SF=16, t4codes) (SF=32, 30codes) 727

You might also like