Performance Effects of the Uplink Asynchronism in a Spread Spectrum Multi-Carrier Multiple Access System

Transcription

1 Special Issue Performance Effects of the Uplink Asynchronism in a Spread Spectrum Multi-Carrier Multiple Access System STEFAN KAISER German Aerospace Center (DLR), Institute for Communications Technology, D-834 Oberpfaffenhofen, Germany Stefan.Kaiser@dlr.de WITOLD A. KRZYMIEN TRLabs/University of Alberta, 8 Park Plaza, Avenue, Edmonton, Alberta, Canada T5K P7 wak@edm.trlabs.ca Abstract. In this paper we investigate the effects of the asynchronism of user signals in the uplink of a spread spectrum multi-carrier multiple access (SS-MC-MA) system. Different propagation delays due to different distances between the mobile transmitters and their base station cause asynchronous arrivals of the uplink signals in addition to multipath propagation. The proposed uplink SS-MC-MA scheme uses for synchronization only the frame structure received on the synchronous downlink and requires no additional synchronization measures. A guard interval that is smaller than the sum of the maximum time offset between the users and the maximum excess delay of the frequency selective multipath channel is used. It minimizes the loss in bandwidth efficiency by allowing residual interference. The residual interference is minimized by proper positioning of the detection interval in the receiver. It is shown that with this approach the guard interval can be reduced by about 5%. The performance of the proposed uplink scheme is compared to an uplink SS-MC-MA scheme with perfect synchronization. 1 INTRODUCTION Future wireless mobile communication systems will have to provide a wide range of multi-media services such as speech, image, and data transmission with variable bit rates. The new services have to be available in different indoor and outdoor application environments. To meet these requirements, it is important to consider recent developments in wireless communications, such as the application of multi-carrier (MC) modulation [1, ] to provide an attractive alternative to the conventional multiple access schemes FDMA, TDMA, and DS-CDMA. MC modulation in the form of orthogonal frequency division multiplexing (OFDM) has already become a key component of the European standards for digital audio broadcasting (DAB) and terrestrial digital video broadcasting (DVB-T). Attention has now focussed on the combination of OFDM with CDMA, TDMA, and FDMA. Most research activity so far has concentrated on MC-CDMA (e.g. [3, 4, 5]). Since MC modulation based approaches require accurate time and frequency synchronization to avoid inter-symbol inter- This work was performed during S. Kaiser's research visit at the TelecommunicationsResearch Laboratories (TRLabs), and was funded by TRLabs, the German Aerospace Center (DLR), and the Natural Sciences & Engineering Research Council (NSERC) of Canada. ference (ISI) and inter-channel interference (ICI), and to achieve high bandwidth efficiency, most of the MC-CDMA schemes have been developed for the synchronous downlink. It has been shown that MC-CDMA is a very promising multiple access scheme for the downlink of a mobile radio system where it enables the deployment of efficient, low complexity receivers employing simple channel estimation [3, 4, 5, 6]. However, this statement does not apply to the uplink, where more complex multi-user detection techniques are necessary to counteract the multiple access interference (MAI), since in the uplink orthogonal spreading codes cannot be used to reduce the MAI. For the synchronous uplink, a promising scheme called spread spectrum multi-carrier multiple access (SS-MC-MA) has been recently proposed [6, 7] to appropriately exploit the advantages of MC-CDMA evident on the downlink. On a synchronous uplink SS-MC-MA is an interesting alternative to other MC multiple access schemes such as MC-FDMA, MC-TDMA, and MC-CDMA [6, 8], as discussed in [9]. There are various possibilities to deal with the time offsets between the user signals in the uplink. The uplink time offsets at the base station can be eliminated as in the GSM system [1] by estimating the time delay between a mobile transmitter and its base station, and correcting for it. Vol. 1, No. 4, July-August

2 S. Kaiser, W. A. Krzymien However, the required synchronization scheme is complex. Another solution is to insert at each OFDM symbol a guard interval (which has to be longer than the sum of the maximum time offset between the user signals and the maximum excess delay of the mobile radio channel) containing a cyclic extension. This approach, however, can result in substantial reduction of bandwidth efficiency (especially in large cells) due to the excessive length of required guard intervals. In this paper, we analyze the effects of user asynchronism in the uplink of SS-MC-MA systems. A simple uplink scheme is considered which synchronizes on the frame structure received on the synchronous downlink. In order to reduce the system complexity, no additional synchronization measures are considered for the uplink. We propose and investigate an approach which allows a certain amount of ISI and ICI, but in return reduces the loss in bandwidth efficiency due to the guard interval. By optimally placing the receiver detection interval, the residual ISI and ICI are minimized, and the complexity of a dedicated uplink synchronization scheme is avoided. The residual interference is evaluated and the resulting performance degradation is determined by comparison with a perfectly synchronized uplink. The paper is organized as follows. In Section the principle of SS-MC-MA in the uplink is described. The loss of synchronism in the uplink due to different signal propagation delays is discussed in Section 3. The resulting interference is described mathematically in Section 4. In Section 5 we present a method to mitigate interference in the uplink, intentionally allowed to reduce the loss of bandwidth efficiency. Performance of the proposed scheme is compared to a perfectly synchronized SS-MC-MA in Section 6. Finally, Section 7 summarizes the results. SS-MC-MA UPLINK STRUCTURE data source of user k channel encoder symbolmapper interleaver The block diagram of the mobile SS-MC-MA transmitter for user k is shown in Fig. 1. After channel coding, code bit interleaving, and QPSK symbol mapping, L complex-valued data symbols d (k) l, l = ::: L ; 1, of user k are spread by multiplication with orthogonal Walsh- Hadamard codes of size L, and superimposed with each other on L subcarriers. The L orthogonal spreading codes are cl =(c l c l 1 ::: c l L;1 ), l = ::: L; 1. The re- serial-to-parallel converter spreader c 1... spreader c L s + (k) serial-to-parallel converter Figure 1: SS-MC-MA transmitter 1 L OFDM with user specific frequency mapper sulting sequence is (k) s =(S (k) S(k) 1 ::: S(k) L;1 ), where... L;1 S (k) j = d (k) l c l j j = ::: L; 1 () l= Variables which can be interpreted as values in the frequency domain, like the elements S (k) j, j = ::: L; 1, each modulating another subcarrier frequency, are written with capital letters. The elements S (k) j, j = ::: L; 1, modulate in parallel the subcarriers assigned to user k. In order to optimally exploit the frequency diversity offered by the mobile radio channel, the subcarriers assigned to different users are interleaved so that the subcarriers used by a given user are spaced by K=T s. This subcarrier assignment is referred to as user specific frequency mapping [6, 7]. To achieve MC modulation/demodulation the OFDM operation is applied. It is efficiently implemented with IFFT/FFT algorithms [1]. A block of N c subcarriers modulated by one set of s (k), k =1 ::: K, is referred to as an OFDM symbol of duration T s. According to the OFDM principles, the N c substreams modulate subcarriers with a spacing of 1=T s. Possible ISI and ICI can be mitigated by inserting a guard interval of duration T g between successive OFDM symbols [, 11]. The guard interval is occupied by a cyclic prefix, resulting in the extended OFDM symbol of duration T s = T g + T s. The block diagram of the SS-MC-MA receiver, located at the base station, is shown in Fig.. After MC demodu- channel The SS-MC-MA system investigated in this paper accommodates K simultaneously active users in the uplink. It is an FDMA scheme on subcarrier level, and each user k, k =1 ::: K, exclusively transmits on a set of L subcarriers out of a total of N c subcarriers. The total number of subcarriers is given by channel inverse OFDM with user specific frequency demapper 1... L parallel-to-serial converter (k) r detector and symbol demapper deinterleaver Figure : SS-MC-MA receiver data sink of user k channel decoder N c = KL: (1) lation with the inverse OFDM operation and deinterleaving (i.e., user specific frequency demapping), the demodulated sequence r (k) =(R (k) R(k) 1 ::: R(k) L;1 ) is obtained. Any conventional, iterative, or joint detection technique suitable for MC-CDMA can be applied for data detection. Joint detection for SS-MC-MA means that L data symbols of one user are jointly detected. We consider a maximum likelihood detector for the detection of the data of user k. After symbol demapping, code bit deinterleaving, and channel decoding, the detected source bits of user k are obtained. 4 ETT

3 Performance Effects of the Uplink Asynchronism in a Spread Spectrum Multi-Carrier Multiple Access System 3 ASYNCHRONISM IN THE UPLINK We consider the uplink of a single cell system. A lowcomplexity synchronization approach is used in which the users synchronize on the frame structure received on the synchronous downlink. In the uplink, significant time offsets between the signals arriving at the base station occur due to different propagation distances between the mobile stations and the base station. The maximum time offset between signals from different users within a cell arriving at the base station is max = R c (3) where R y is the radius of the cell and c is the speed of light. The factor results from the summation of the propagation delays in the downlink and the uplink. The delays of the signals of the K users due to the propagation distance to the base station are (k) [ max ], k =1 ::: K. With the assumption of uniform surface distribution of users within the cell, the resulting linear probability density function of the signal delays at the base station is p ( (k) )= 8 < : (k) max if (k) max otherwise : (4) As expected, long delays from users at the cell boundary occur most often. Without compensation of different propagation delays of the signals in the uplink and an insufficient guard interval, the user synchronism is lost and interference results, which can significantly deteriorate the performance of an OFDM system. In the following sections, the resulting interference is described and evaluated. 4 INTER-SYMBOL AND INTER-CHANNEL INTERFERENCE In this section we present a general description of ISI and ICI in an asynchronous multi-carrier link. We assign each subchannel n its own delay n, n = ::: N c ; 1. The delays on subcarriers assigned to the same user are determined by the location of the user, and are the same. One arbitrary symbol S n i is transmitted per subcarrier in one OFDM symbol. The index n is the subchannel index and i is the OFDM symbol index (discrete time). The symbols S n i are equivalent to S (k) j with their dependence on discrete time explicitly shown. In the sequel, the notation S n i is preferred to describe the ISI and ICI for the sake of clarity. There are N c orthogonal subcarrier frequencies f n, separated from each other by 1=T s. The nth orthogonal basis function is defined as e jf nt for ; T g t<t s n (t) = (5) otherwise y When the cells are not circular, R is the maximum distance between the base station and a mobile user belonging to a given cell. and the signal transmitted on subcarrier n is given in time domain by x n (t) = 1 i=;1 S n i n (t ; it s ; n) T s = T s + T g : (6) Considering multipath propagation with N p paths, the impulse response of the wideband mobile radio channel associated with the nth subcarrier is given by h n (t) = N p;1 n p (t ; n p ) (7) where n p is the complex-valued attenuation. The delay n p represents the excess delay of the pth path of subchannel n. The transmission bandwidth, 1=T s, used by the nth subchannel is small compared to the coherence bandwidth of the mobile radio channel associated with the nth subcarrier and described by h n (t). Therefore, fading on the nth subchannel of bandwidth 1=T s is flat [1]. For the sake of generality, an impulse response h n (t) is defined for each subchannel. For simplicity it is assumed that all channel impulse responses h n (t), n = ::: N c ;1, have the same number of paths N p. The total delay n p per arriving path p on subchannel n at the base station is the sum of the propagation delay n and path delay n p, i.e., n p = n + n p (8) where n p [ max ], max = max + max, and max is the maximum excess delay of any subchannel. The received signal y(t) is the sum of the convolutions of x n (t) with h n (t) over all subcarriers and an additive noise term n(t), thus, y(t) = N c;1 (x n (t) h n (t)) + n(t) (9) where the symbol denotes convolution. The demodulation of subcarrier m in the receiver involves a simple correlation with m (t ; it s ) over the effective OFDM symbol duration T s, i.e., R m i = 1 T s Z T s y(t) m (t ; it s ) dt = Y m i + N m i : (1) The symbol (:) denotes complex conjugation. The component N m i corresponds to the additive noise component n(t) after demodulation. The output Y m i of the demodulator is the noise-free decision variable that includes ISI and ICI. Y m i can be written as Y m i = 1 T s Z Ts n p xn(t ; n p) m (t ; its ) dt (11) Vol. 1, No. 4, July-August

4 S. Kaiser, W. A. Krzymien = = = where 8 >< >: 1 T s m (t ; it s ) dt n p Z Ts 1 i =;1 S n i n(t ; i T s ; n p) n p (Sn i m n(n p) +Sn i;1 m n(n p)) h m p (Sm i m m(m p) +Sm i;1 m m(m p)) + n6=m n p (Sn i m n(n p) +Sn i;1 m n(n p)) i m n(n p) = (1) and if n p <T g sin ((n ; m)( n p ; Tg )=Ts) (n ; m) if n p Tg e j(n(t s ;n p);(n;m)(n p;tg))=ts m n(n p) = (13) 8 >< >: sin((n ; m)) (n ; m) sin ((n ; m)(ts ; n p)=ts) (n ; m) e j(nn p;(n;m)(t s ;n p))=t s if if n p <T g n p Tg The desired component in (11) on subcarrier m is m p S m i m m ( m p ) and the ISI on the same subcarrier is m p S m i;1 m m ( m p ). ICI from the subcarriers n 6= m is given by n p (S n i m n ( n p )+S n i;1 m n ( n p )). The average signal-to-interference power ratio (SIR) on subcarrier m is P s (SIR) m = (14) P ICI + P ISI where P s is the power of the desired component and P ICI + P ISI is the interference power. With the assumption that the average received power on all subcarriers is the same due to power control and, furthermore, that the transmitted data symbols are statistically independent, the average SIR on subcarrier m is (SIR)m = (15) N P p;1 E j m p m m (m p)j P ; P E j n p m n(n p)j + P j n p m n(n p)j n6=m where Ef:g is the expectation. 5 INTERFERENCE MITIGATION : Two detection concepts differing in the choice of the position of the detection interval in the receiver are investigated. The SS-MC-MA uplink scheme which minimizes the residual interference is referred to as Concept I. In the design referred to as Concept II each user minimizes the interference originating from its own set of subcarriers, but it does not attempt to minimize the interference from other asynchronous users. In other words, in Concept II each user optimizes its detection interval with respect to its own multipath channel. For an SS-MC-MA mobile radio system with channel estimation as described in [6, 7], the assumption can be made that the base station has the knowledge about the arrival time (k) of each user. To obtain (k) simply a correlation of the received signal with the known pilot symbol grid on the subcarriers exclusively used by a given user has to be performed. Moreover, we assume to know the maximum excess delay max on the radio channels used. A given value of max can be assumed in the uplink design, or its actual value can be obtained from channel estimation. The subset of subcarriers exclusively assigned in SS- MC-MA to user k is denoted as A (k). When all subcarriers are in use, we obtain N c = K k=1 jja (k) jj (16) where jja (k) jj is the number of elements in subset A (k). 5.1 CONCEPT I In order to keep the loss in bandwidth efficiency due to the insertion of guard intervals at a tolerable level, we consider an uplink scheme which allows ISI and ICI by choosing T g < max, i.e., we use a short guard interval. At the same time the detection interval in the receiver is chosen such that the residual interference is minimized. Principle features of the Concept I are illustrated in Fig. 3. To calculate the SIR for Concept I according to (15) and using (1) and (13), a new time delay variable has to be defined. Signals arriving earlier compared to the detection interval have to be transformed to the case with a delayed arrival since (1) and (13) are only valid for delays n p, i.e., not for earlier arriving signals. The new time delay variable n p is defined as n p = n p ; ; n p + T g if n p otherwise (17) where +T g is the beginning of the integration interval for demodulation (of duration T s ), which is the same for all users. The time shift can take on values in the interval [ max ; T g ). The case n p in (17) describes the situation where the signal on subcarrier n is delayed compared to the detection interval. The operation n p ; shifts the zero of the time axis such that (1) and (13) can be applied. The case n p < in (17) is valid when the signal on subcarrier n arrives earlier compared to the detection 4 ETT

5 Performance Effects of the Uplink Asynchronism in a Spread Spectrum Multi-Carrier Multiple Access System p( ε) ε detection interval max < max + max, and otherwise (see Fig. 3). The interference which is not absorbed by the guard interval is minimized by choosing as guard interval OFDM symbol δ max ε max ε = max + max ; T g 1+max () = max { Concept I { Concept II +T g δ (k-1) T s δ (k-1) +T g T s user k user k-1 user k+1 user k-1 δ (k-1) user k user k+1 Figure 3: Principle of interference mitigation with Concept I and Concept II. interval. The operation ; n p inverts the earlier arrivals to delays such that (1) and (13) can be applied. The additional shift of T g is necessary since interference occurs always with earlier arrived signals and no advantage can be gained from the guard interval, i.e., only the second case in (1) and (13) is possible. The transformation of earlier arrivals to delays is valid since the interference power is the same due to symmetry. With Concept I and the new time delay variable n p the SIR is given by (SIR) (k) = E + E ( ( 8mA (k) 8mA (k) j n p m n(n p)j j m p m m (m p)j 8nA (g) 8g6=k )! ;1 ) j n p m n(n p)j (18) Through (17) the SIR depends on the parameter. To optimally choose, the probability density function p () of the total delay n p, n = ::: N c ; 1, p = ::: N p ; 1, has to be considered. The probability density function p () is obtained by convolution of the probability density function p ( (k) ), see (4), and the probability density function p ( ) of the multipath propagation delay p () =p () p (): (19) Given (4) and a p ( ) as an exponentially or linearly decreasing function with maximum excess delay max, the probability density function p () can be approximated by a linearly increasing function in the interval < max and linearly decreasing in the interval given the approximation of p () by a triangle. Numerical evaluations of with the exact p () showed that the approximation used in this paper resulting in the analytical solution () is sufficiently accurate. Without multipath propagation, the integration interval for demodulation of the data of all users starts at the time instant max. In that case, all interference occurring from users with delays in the interval [ max ; T g max ) is cancelled out. Thus, ICI is minimized due to the linearly increasing characteristic of (4). For Concept I we can define the worst case user as the user which is next to the base station and has a delay of (k) s. Due to the fixed detection interval starting at +T g, the worst case user in addition to the ICI from other users suffers from the maximum possible ISI. 5. CONCEPT II Concept II is the straightforward solution, where the integration interval for demodulation of the data of user k starts at the time instant (k) + T g. The integration interval can thus be different for each user in order to minimize the interference originating from its own subset of subcarriers. Hence, this approach does not take into account the probability density function p ( (k) ) of the delays and, thus, of the interference from other asynchronous users. For Concept II the SIR given by (18) reduces to (SIR) (k) = E E ( ( 8mA (k) 8mA (k) 8nA (g) 8g6=k j m p m m(m p)j ) j n p m n(n p)j +j n p m n(n p)j )! ;1 (1) when T g > max, since as long as T g > max, this approach avoids ISI, even if T g < max. Principle features of the Concept II are illustrated in Fig. 3. For Concept II we can define the worst case user as the user with maximum distance R to the base station and a maximum delay of (k) = max. Thus, each user with (k) < max interferes with the desired user, and the guard interval gives no advantage. Vol. 1, No. 4, July-August

6 S. Kaiser, W. A. Krzymien 6 SIMULATION RESULTS The asynchronous uplink is investigated for two different SS-MC-MA mobile radio systems. System A is designed for medium size cells with a radius of about km, typical for future outdoor cellular mobile radio systems. System B is designed for high data rate indoor and microcellular outdoor applications with cell radius up to 5 m. Table 1 gives the parameters of both systems. Table 1: Parameters of the SS-MC-MA systems System A System B environment outdoor indoor cell radius R km 5 m max. delay max s 1.67 s bandwidth B MHz MHz num. of subcar. N c OFDM sym. dur. T s 18 s 1.4 s guard duration T g 15 s 1 s user capacity K sys net bit rate R b.11 to 1.4 Mbit/s.13 to 15.8 Mbit/s carrier frequency f c GHz spreading code Walsh-Hadamard code of length L =8 symbol mapping QPSK channel code convolutional code of rate 1/ and memory 6 The cell radius R and the guard interval duration T g are varied in later figures to demonstrate the effect of cell size on performance. The channel estimation is assumed to be perfect. However, in the calculation of the net bit rate per user a % loss due to pilot symbols is taken into account. Both systems use a TDMA frame structure where N fr is the number of time slots per TDMA frame, explained in detail in [6, 7]. The duration of one TDMA frame is about 18 ms. In System A one time slot contains 31 OFDM symbols, and N fr = 4 time slots form a TDMA frame. In System B one time slot contains 9 OFDM symbols, and one TDMA frame has N fr =6time slots. The difference in the number of time slots between System A and B results from different OFDM symbol times, T s, and guard times, T g, for the two systems. The user capacity of the system is K sys = N fr K: () The total bit rate is the bandwidth B times for QPSK and divided by due to the 1/ channel code rate. Moreover, the % loss due to pilot symbols and the loss due to the guard interval have to be taken into account. The net bit rate per user R b is obtained by dividing the total bit rate by K sys which results in 11. kbit/s for System A and 1.3 kbit/s for System B. It is possible to assign to one user several or all transmission links, such that a net bit rate up to 1.4 Mbit/s for System A, and 15.8 Mbit/s for system B is obtained. The mobile radio channel models are taken from [13]. For System A the 'Outdoor Residential -High Antenna' (Channel B) channel model with maximum excess delay max = 15 s is chosen. Velocity of the mobile user in System A is 3 km/h, resulting in the maximum Doppler frequency of Hz, and the classical Doppler spectrum is assumed [13, 14]. For System B the 'Indoor Commercial' (Channel B) channel model with maximum excess delay max = 75 ns is used. In this case the velocity of the mobile user is 3 km/h, the resulting Doppler frequency is 5.56 Hz, and the flat Doppler spectrum is assumed [13]. Thus, System A results in the maximum delay of max = 8:33 s, and System B of max = :4 s. The results shown in the following are obtained through Monte Carlo simulations. It has to be mentioned that for the case of an additive white Gaussian noise (AWGN) channel or a flat Rayleigh fading channel analytical solutions exist to evaluate the uplink performance of asynchronous SS-MC-MA. All simulation results shown in the following are for the most critical case of a fully loaded system. Moreover, the signal-to-noise ratio (SNR) degradation due to the guard interval and pilot symbols (%) is taken into account in the presented results. In Figs. 4 and 5 the SIR versus the guard time T g is depicted for System A and System B, respectively. The SIR SIR in db Concept I, R=1km Concept I, R=km Concept I, R=3km Concept II, R=1km Concept II, R=km Concept II, R=3km T g in µs Figure 4: SIR for System A for different cell sizes (worst case user, fully loaded system). is shown with interference mitigation according to Concept I and Concept II for different cell sizes. The worst case user of Concept I and Concept II is considered. It is obvious that only Concept I is able to deal with the interference since the straightforward solution of Concept II does not take into account interference from asynchronous users. The BER versus the SNR (E b =N ) for the uncoded System A is shown in Fig. 6, and for the uncoded System B in Fig. 7. The results are presented for the interference mitigation Concept I and Concept II, with the perfect synchronization case (no ISI, no ICI) as a reference. For System A with cell radius R =km, even for the worst case user considered for Concept I, the degradation compared to the 44 ETT

7 Performance Effects of the Uplink Asynchronism in a Spread Spectrum Multi-Carrier Multiple Access System SIR in db perfect synchronization case is less than 1 db between the BERs of 1 ; to 1 ;3. Thus, for System A with Concept I the guard interval can be reduced by about 5% of max. For System B the reduction of the guard time compared to max can be even higher. Finally, Fig. 8 presents the BER versus the SNR for a coded SS-MC-MA system (System A). This time only Concept I, R=5m Concept I, R=5m Concept II, R=5m Concept II, R= 5m T g in µs Figure 5: SIR for System B for different cell sizes (worst case user, fully loaded system). BER no ISI, no ICI Concept I, R=1km Concept I, R=km Concept I, R=3km E b /N in db BER 1 3 no ISI, no ICI Concept I, R=1km Concept I, R=km Concept I, R=3km Concept II, R=1km Concept II, R=km Concept II, R=3km Figure 8: BER performance of System A with channel coding (rate 1/ code, worst case user, fully loaded system). Concept I is considered. For the worst case user the degradation is about.5 db at a BER of 1 ;4 and cell radius R =km E b /N in db Figure 6: BER performance of the uncoded System A (worst case user, fully loaded system). BER no ISI, no ICI Concept I, R=5m Concept I, R=5m Concept II, R=5m Concept II, R=5m E b /N in db Figure 7: BER performance of the uncoded System B (worst case user, fully loaded system). 7 CONCLUSIONS In this paper we have investigated the uplink performance of an asynchronous spread spectrum multi-carrier multiple access (SS-MC-MA) system. A simple uplink scheme is proposed which synchronizes only on the frame structure received in the synchronous downlink. By allowing residual interference, the guard interval can be reduced to a value ensuring high bandwidth efficiency of the system. Simulation results have been obtained for various uncoded and coded SS-MC-MA systems in mobile radio cells of various sizes. These results are compared to a perfectly synchronized SS-MC-MA system. By proper positioning of the detection interval, the otherwise necessary guard interval can be reduced by about 5%, resulting in increased bandwidth efficiency and only moderate SNR degradation on the order of.5 db. Manuscript received on January 9, 1999 Vol. 1, No. 4, July-August

8 S. Kaiser, W. A. Krzymien REFERENCES [1] S. Weinstein and P. M. Ebert, Data transmission by frequency-division multiplexing using the discrete Fourier transform, IEEE Trans. Commun. Tech., vol. 19, pp , Oct [] J. A. Bingham, Multicarrier modulation for data transmission: An idea whose time has come, IEEE Commun. Mag., pp. 5 14, May 199. [3] K. Fazel, S. Kaiser, and M. Schnell, A flexible and high performance cellular mobile communications system based on orthogonal multi-carrier SSMA, Wireless Personal Commun., vol., no. 1&, pp , [4] N. Yee, J.-P. Linnarz, and G. Fettweis, Multi-carrier CDMA in indoor wireless radio networks, IEICE Trans. Commun., vol. E77-B, pp. 9 94, July [5] L. Tomba and W. A. Krzymien, Downlink detection schemes for MC-CDMA systems in indoor environments, IEICE Trans. Commun., vol. E79-B, pp , Sept [6] S. Kaiser, Multi-Carrier CDMA Mobile Radio Systems Analysis and Optimization of Detection, Decoding, and Channel Estimation. Dusseldorf: VDI-Verlag, Fortschrittberichte VDI, series 1, no. 531, 1998, Ph.D. thesis. [7] S. Kaiser and K. Fazel, A flexible spread-spectrum multicarrier multiple-access system for multi-media applications, in Proc. IEEE Int. Symp. on Personal, Indoor and Mobile Radio Commun. (PIMRC'97), pp. 1 14, Sept [8] H. Rohling and R. Gruenheid, Performance comparison of different multiple access schemes for the downlink of an OFDM communication system, in Proc. IEEE Vehic. Technol. Conf. (VTC'97), pp , May [9] S. Kaiser, MC-FDMA and MC-TDMA versus MC-CDMA and SS-MC-MA: Performance evaluation for fading channels, in Proc. IEEE Fifth Int. Symp. on Spread Spectrum Techniques & Applications (ISSSTA'98), pp. 4, Sept [1] S. M. Redl, M. K. Weber, and M. W. Oliphant, An Introduction to GSM. Artech House Publishers, [11] E. Viterbo and K. Fazel, How to combat long echoes in OFDM transmission schemes: Sub-channel equalization or more powerful channel coding, in Proc. IEEE Global Telecommun. Conf. (GLOBECOM'95), pp , Nov [1] J. G. Proakis, Digital Communications. McGraw-Hill, [13] K. Palavan and A. H. Levesque, Wireless Information Networks. John Wiley & Sons, [14] W. C. Jakes, Microwave Mobile Communications. John Wiley & Sons, ETT