3216

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

A Time Reversal Transmission Approach for Multiuser UWB Communications

Hung Tuan Nguyen, Student Member, IEEE, Istvn Zsolt Kovcs, Member, IEEE, and Patrick Claus Friedrich Eggers, Member, IEEE
AbstractWe propose and evaluate the performance of the time reversal technique in impulse radio ultrawideband (UWB) communications. The evaluation was based on measured channel impulse responses of a 4 1 multiple input single output (MISO) system in the UWB frequency bandwidth of 3 to 5 GHz with both vertical and horizontal polarization at the receiver. The results show that there is a great potential in combining time reversal and UWB technique with respect to both reducing the receiver complexity and improving the system performance. Simultaneous communication is illustrated with 5 users, each with a bit error rate (BER) of less than 10 3 at an average signal to noise ratio (SNR) of 15 dB. Index TermsMultiple elements antenna, multiple input single output (MISO) system, spatial focusing, time reversal (TR) techniques, ultrawideband (UWB).

I. INTRODUCTION LTRAWIDEBAND (UWB) is among the hot topics today due to its potential in short range communications and positioning applications. With UWB, the dense multipath components can be resolved allowing the use of Rake receiver for signal demodulation [1], [2]. UWB aims at providing communications with high bit rates and low complexity devices using an extremely wide frequency range. Time reversal (TR) techniques have been applied extensively in acoustic, medical and under water communications applications [3], [4]. In [4], it was demonstrated that in the ultrasonic frequency regime it is possible to provide error free communication with 5 receivers simultaneously. Because of its simplicity and performance advantages, the idea of applying the TR in wireless communication has gained much attention recently. It is expected that TR can reduce the effect of inter symbol interference (ISI) signicantly without the need of a high complexity equalizer at the receiver. Evaluation on the performance of TR based on wireless channel measurements in [5][7] and [9] among others has shown its great potential. Recently the authors in [8] have proposed an approach using TR with a MMSE equalizer and applied it to UWB communications. The use of TR has therein been illustrated as a promising solution in single user UWB communications. In this work we address the applicability of TR in impulse radio UWB transmissions as one possible solution to improve

the multiuser system capacity and communication range. It is expected that the extremely wide bandwidth of the UWB signal will help in increasing the temporal compression and in improving the spatial focusing characteristics of TR. On the other hand, TR may help in reducing the complexity of the receiver devices and in lowering the required transmitting power in an UWB system. In our proposed scheme, by using only a single tap which contains most of the signal energy the receiver does not need to estimate the channel. Because no equalizer is used this can result in an easier synchronisation process at the receiver and the complexity of the receiver can be reduced signicantly. These are also the major differences in our approach compared to the approach proposed in [8]. We show that under certain assumptions, TR-UWB can fulll the receivers low complexity requirement as well as increase the system capacity in terms of the number of supported users. The performance of the proposed scheme was evaluated using radio channel data from recent MIMO UWB channel investigations [12]. The rest of the paper is organized in the following way. First we briey review the TR technique and present the proposed TR-UWB transmission scheme in Section II. The UWB measurement setup and scenarios are presented in Section III. Based on the measured channel impulse response (IRs) the TR-UWB performance in terms of the signal to interference ratio (SIR), the number of simultaneously supported users, and the system BER are analyzed in Section IV. The conclusion wraps up the paper in Section V. II. MULTI ELEMENT ANTENNAS AND TR A. TR Back Ground and Performance Metrics The essence of the TR technique is that by coherently adding the received signals in the delay domain and the spatial domain (multiple antennas) we can achieve both the time compression and the spatial focusing at the same time. Due to the spatial focusing the interference can be reduced and due to the time compression the ISI can be alleviated. The mathematical formulation of the received signal of a SISO link using TR in the noise free situation can be described as follows: (1)

Manuscript received September 12, 2005; revised March 12, 2006. This work was supported in part by an Aalborg University Ph.D. Fellowship Grant and in part by the IST 2004-507102 MAGNET project. H. T. Nguyen and P. C. F. Eggers are with the Department of Communication Technology, Aalborg University, DK-9220 Aalborg, Denmark (e-mail: [email protected]; [email protected]). I. Z. Kovcs is with the Network Systems Research/Nokia Networks, DK 9220 Aalborg, Denmark (e-mail: [email protected]). Color version of Fig. 3 is available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2006.883959

where denotes convolution, indicates the complex conis the autocorrelation of the channel IR jugate; between th transmitting antenna and th receiving antenna; is the transmitted symbol. The received signal at any off-target point has the form (2)

0018-926X/$20.00 2006 IEEE

NGUYEN et al.: A TIME REVERSAL TRANSMISSION APPROACH

3217

where denotes the IR of the channel from the transmitting We note that in order to apply TR, the channel IR must be is the cross correlation known at the transmit side. In a reciprocal channel, the channel point to the off-target point and and the time reversed complex conjugated estimationofthebackwardlink(fromthereceivertothetrans- thay the of the channel IR . version of the transmitted signal mitter) can be thought of as a solution to estimate the IR of The three fundamental propagation mechanisms: reection, the forward link (from the transmitter to the receiver). This rediffraction and scattering, which in general make the propa- quirement on reciprocal channel can be approximately fullled gation channel of each communication link independent and in a systems employing time-division duplexing (TDD) in a unique, contribute to the process of magnifying the intended quasi-stationary environment. signal and attenuating the interference at the target point. It turns out that the environment can provide the articial orthogonal B. TR-UWB Transmission Scheme code and that the more decorrelated the channel the better the In impulse radio UWB technique, the transmitted pulse width orthogonality property of the code. is much smaller than the effective length of the measured IR. TR for a SISO system works in the same manner as the tra- Therefore the convolution of the signal with the time-reversed transmitting an- IR is essentially equivalent to the multiplication of time-reditional pre-Rake system. When there are tennas, under ideal conditions the received signal will be the versed IRs tap with the transmitted UWB pulse. As long as the IR autocorrelations. sum of pulse width is smaller than the interval between two taps, the real-valued pulses between two consecutive taps are distinguishable, i.e., The received signal peak is the sum of the correlated numbers, while the side lobe signals are the sum of there is no interference between them. Therefore the spectral real-valued random numbers. This gives rise to the spatial characteristics of the transmitted TR waveforms are essentially focusing in systems using TR. determined by the spectral characteristics of the transmitted antennas pulse waveform. The receiver only needs to synchronize with Now we consider the scenario when there are antennas at the receive side. The the main peak of the received signal. It does not need to estiat the transmit side and received signal at the th receiving antenna is mate the channel IR and can use a single tap/Rake nger tuned to the major peak. Thus the detection of the received signal can be done with a rather simple energy detection method. Because of the temporal focusing characteristics of the TR, it is expected that synchronization to the main peak of the signal at the receiver would be easy. Since no equalizer is required, the design of the receiver becomes fairly simple. Most of the complexity burden is moved to the transmitter and thereby the cost and the size of the receiver can be reduced. Performance evaluations of the pre-RAKE scheme for SISO UWB system (3) (TR-UWB for SISO) can be found in [10], [11] among others. For illustration purposes, initially the interval between two transmitted symbols is assumed to be equal to the length of the and are the transmitted symbols intended for where measured and estimated channel IRs. The ISI effect therefore the th user/receiving antenna and for the th user/receiving ancan be neglected. However, because of the TRs temporal fotenna respectively. In the following we use the term equivalent cusing characteristics, the transmitted symbol interval can be to denote the sum of the autocorrelation in (3) IR or smaller than the effective IRs length as long as the ISI effect is (4) not severe. Fig. 1 illustrates the working principle of a TR-UWB MISO system. Spatial focusing is the other main advantage of applying TR As long as the channels are uncorrelated, we expect that the in UWB communications. Because the transmitted signal is fointerference part in (3) can be suppressed. Thereby, it might cused in signal space on the target user, not only is the interbe possible to communicate with all users simultaneously with ference to other users reduced but so is the interference to other a simple detection scheme at each receiver. Consequently, wireless communication system. In the context of multiuser scethe throughput of the system can be signicantly increased. nario, the system capacity therefore increases (see Section IV-D It should be noted that the formulation mentioned above can for detail). MIMO system or be used to describe a single simultaneously operating MISO systems, each one C. Equal Power Control targeting a different user. As in [7], we use the Signal to InterThe plain TR transmission technique described in ference ratio (SIR) as a metric representing the spatial focusing Section II-B assumes that the transmitted power of each characteristics of TR. The instantaneous SIR is calculated as symbol for each transmission link is equal to the power of the the ratio of the peak power of the signal of interest and the estimated channel IR. It means that the plain TR transmission corresponding interference power at the same time lag. technique also introduces a sort of power control for each indi(5) vidual transmission link. However when there are simultaneous communication links, due to the difference in the transmitted power intended for different receiving antennas/users, the where denotes the absolute value operation.

3218

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

Fig. 1. Illustration of a practical MISO TR-UWB single user system of transmitting antennas and one receiving antenna with real IRs taken from measurement. Fig. 2. Signal applied at one antenna for MISO TR-UWB multiple-user scenario: normal TR case (left), proposed shifted TR (TR&S) scheme (right). The IRs were real IRs taken from measurement.

interference power [as described in (3)] can be comparable to (or a large fraction of) the intended signal. Here we use the same approach described in [7] as a solution. The time-reversed IR is normalized by the measured wideband power so that the transmitted power intended for each receiving antenna/user will be equal (6) where denotes the Frobenius norm operation.

D. Equal Power Control & Interference Suppression For a TR system with a large number of transmitting antennas compared to the number of simultaneously receiving antennas/ users the equal power control solution (Section II-C) can be used to reduce the effect of branch power difference. And thereby the interference can be suppressed. However, in practice the number of transmitting antennas at a hot spot or base station cannot be innitively large because of the constrains in the size, cost etc. The expectation of simultaneous transmission of independent data stream requires that the time-reversed IRs should be aligned at their maximal peak/or the rst arrival path, so that one can independent symbol simultaneously Fig. 1. Betransmit cause of the time alignment, even though the channel IRs are uncorrelated, the magnitude of the IRs cross-correlation might still be an order of magnitude smaller than that of the IR . In general, the interference at the peak autocorrelation of the signal also increases linearly with the number of simultaneous transmissions. Low transmission quality, i.e., low SIR is traded for more users. Our proposal for interference suppression consists in offsetting in time the time-reversed IRs intended for each user. The transmitted signal intended for th user at the th transmitting antenna can be described as

resolution. Fig. 2 illustrates the signals transmitted at one transmitting antenna before and after shifting the time-reversed IRs taps relative to each other. main peaks by In this way the taps in the propagation channel containing signicant energy are multiplied by taps of the transmitted signal with low energy and vice versa. The peak to peak multiplication can be avoided and the interference power is signicantly , suppressed. Depending on the amount of the time offset the system capacity is reduced as it requires a longer interval for all symbols to be transmitted. The communication quality is improved as the interference is decreased. The results of the SIR as a function of the time offsetting interval are shown in the Section IV. Hereafter we use the term TR&S to denote the shifted time reversed scheme. III. MEASUREMENT AND DATA ANALYSIS The channel measurements were conducted with a UWB sliding correlator channel sounder. The measurement bandwidth was 2.5 GHz centered at 4.5 GHz. The effective delay ns. The antennas used were planar resolution was monopole antennas with a bandwidth of 2 GHz to 15 GHz. An array of four elements was used at the access-point (AP) device, while a single element was used at the user terminal with either vertical or horizontal polarization, (see Fig. 3). The environment was quasi-static during the measurements and only the stationary user device scenarios were used for the results presented in this paper. The environment was a large hall (17 m 12 m 11 m) with two metalised glass walls, two metallic bridges and one staircase structure. The AP antenna system was mounted at 6 m height above the oor. The user terminal was moved to ve different locations always at a height of 0.8 m above the oor. The measured locations were distributed over an area of 17 12 meters square. The distance between any two measured locations was at most 10 meters. The effective TX-RX distances were in the range of 6 to 17 m. For one type of receiving antenna polarization, at one location 1024 4 (1024 is the number of SISO channel IRs taken over

(7) where denotes the circular shift operation, is the shifted unit calculated as the number of taps and is the tap

NGUYEN et al.: A TIME REVERSAL TRANSMISSION APPROACH

3219

Fig. 4. Comparison of the RMS delay spread CDF, the real measured channel IR (top), the equivalent channel IR (bottom).

Fig. 3. User device with vertical or horizontal polarized element. (a) Aligned time-reversed IRs, simple TR. (b) AP linear antenna array with vertical polarized elements.

time, and is the number of antennas at the AP) channel recorded IRs were captured. The number of equivalent IR, therefore was 1024. The channel at one measured location, IRs were sampled at a rate of 40 Hz. A detailed description of the equipment and investigated environment can be found in [12]. The obtained channel data was compensated for all system components. In order to simulate the impulse radio UWB system, the measured complex IRs were converted into real-valued IRs via standard Hermitian transformation. To illustrate the working principle of TR in UWB, in the following analysis we assumed that the measured data is noise free and the forward link can be perfectly estimated from the backward link (using the reciprocity theorem). IV. RESULTS A. Temporal Focusing From the measurement data we estimated the length of 90% energy window dened as % % (8)

is the length, as where PDP is the power delay prole, a number in taps, of the measured IR. The average value of is about 100 taps or 40 ns. This 90% energy window gives us an idea on the length of the truncated channel IR which should be taken from the measured one. By using , it is believed that most of the information of the measured channel IR will be retained while keeping the truncated channel IR as short as possible. Therefore, we use a truncated version of the measured taps starting from the rst arrival path, IR with or the main peak of the IR in most case, as the effective length of the IR. To illustrate the time compression characteristics of the TR, we compare the instantaneous RMS delay spreads of the origof a 4 1 MISO inal channel IRs and the equivalent IR, system. For the measured IRs, the instantaneous RMS delay spread is calculated from the rst and second moment of the instantaneous IRs

(9) where is the excess time delay. The RMS delay spread of the is calculated by IR

(10) The cumulative distribution function (CDF) of the RMS delay are shown in Fig. 4. spread for the measured IR and the Using TR, the median RMS delay spread of was reduced by a factor of 1.5 compared to that of the original channel IR. This is in agreement with the results shown in [5] and [7].

3220

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

TABLE I MINIMUM, MEAN AND STANDARD DEVIATION VALUE OF THE SSR IN DECIBELS VERSUS THE USED IRS LENGTH FOR VERTICAL ( ) AND HORIZONTAL ( ) POLARIZATION

B. Signal Detection As mentioned in Section II-B, for the TR-UWB scheme to work correctly the receiver needs to be synchronized to the received signal peaks. Due to the temporal focusing property of TR described in Section IV-A, more signal energy can be collected and therefore, this could ease the synchronization procedure. Simple energy threshold detectors [13] or more complex ones, such as the chip-level post-detection integration schemes [14] can be potentially used to achieve synchronization. We calculated the signal to side lobe ratio (SSR), dened as the power ratio of the main signal peak to the largest side lobe as a function of the used IRs signal of the equivalent IR, length, .

Fig. 5. Mean SIR as a function of the time offset and the number of users.

SIR can be obtained, i.e., the median SIR increases to 23 dB from 12 dB for the scenario of 5 simultaneous users. These results become more signicant if we consider the separation distance between users/measurement locations which was only 10 meters at most [12]. For a 2-user scenario, both with (11) UWB signals, the mean SIR of the aligned IRs is about 20 dB. For the same value of SIR, with wideband (UMTS) signals the required separation distance is about 300 m [7]. This demonThe results are shown in Table I. The mean SSR values do not strates a much better spatial focusing performance of TR for very much depend on the length of the used IRs. A slightly lower UWB bandwidth (GHz) than for the UMTS bandwidth (MHz). mean and min SSR values was observed when the receiving anFrom Fig. 6 we can clearly see the effect of multiple users tenna is horizontally polarized. However, the standard deviation to the SIR values. Fewer users means lower interference and of the SIR for this type of receiving antenna is smaller. With therefore higher SIR value can be obtained. a mean SSR of more than 10 dB, the synchronization to the With the aligned IRs scheme, the SIR was slightly higher main received signal peak can be fairly easily achieved. We also when vertically polarized receiving antennas are used calculated the SSR of the original channel IRs. The minimum, [Fig. 6(a)]. Using TR&S scheme, with two users, vertically pomean and standard deviation values were 0, 5.1, and 3.5 dB, re- larized antenna gives a slightly better SIR. However, for more spectively. These numbers are lower than the ones tabulated in than two users, the SIR of the TR system with horizontally Table I. By comparing the SSR values of these two cases, again polarized antenna is higher [Fig. 6(b)]. the temporal focussing characteristics of the TR are illustrated. 2) Effect of Outdated Channel Information: One of the most important conditions for TR to work properly is that the channel C. Spatial Selectivity and TR with Time Offset Scheme IRs must be known at the transmitter. As mentioned earlier, this 1) Results in an Ideal Case: As described in Section II, by can be fullled by estimating the channel from the backward applying the TR&S scheme, the interference at the peak of the link (from the receiver to the transmitter). However, even though can be further suppressed. Fig. 5 shows the mean SIR values the measured channel is quasi-stationary there is concern about as a function of the time offset interval . It is observed that the impact of outdated IRs on the performance of TR. For extaps the SIR saturates ample, the channel can vary in time due to the movement of above a certain offsetting value of at the average value of 27 dB. the people or objects in the vicinity in the measurement site. Fig. 6 shows the CDF of the SIR calculated from the mea- To evaluate this impact, we calculated the SIR value when the sured IRs with the effective IR length of 100 taps (40 ns). The time reversed version of the current channel IR is used as the TR results for both aligned time-reversed IRs and TR&S schemes lter for the next channel realization. Since the channel IR was are given. Note that the offset time described in Fig. 2 can be sampled at the rate of 40 Hz, the delay time therefore is 2.5 ms made dynamic or static depending on the channel condition and and we believe that this is a realistic interval for the loopback system requirement. Henceforth, for simplicity purposes we use delay. The mean and standard deviation of the SIR values for taps, which is equivalent to 2 ns. the perfect and outdated cases are tabulated in Tables II and III. a static offset time of It is observed that by shifting the transmitted time-reversed IRs In most cases, the loss in the mean SIR value caused by outintended for each user by 2 ns, a signicant improvement in the dated TR lter is about 2 dB, which is a reasonable number if

NGUYEN et al.: A TIME REVERSAL TRANSMISSION APPROACH

3221

TABLE II COMPARISON OF THE MEAN/STANDARD DEVIATION VALUES = OF THE SIR IN DECIBELS WHEN USING INSTANTANEOUS AND OUTDATED TR FILTER, VERTICAL POLARIZATION

TABLE III COMPARISON OF THE MEAN/STANDARD DEVIATION VALUES = OF THE SIR IN DECIBELS WHEN USING INSTANTANEOUS AND OUTDATED TR FILTER, HORIZONTAL POLARIZATION

The equal power control scheme, (Section II-C), is used in the simulations. By varying the noise variance, we can vary the synthetic signal to noise ratio (SNR) at each measured location.

(13) is the noise variance. To assure that this SNR value where is always lower than the measured SNR of 25 to 30 dB [12], we only simulate the BER with the SNR range of 0 to 20 dB. For these simulations, we used the bipolar pulse amplitude modulation (BPAM) with the Gaussian pulse shape similar to the investigations in [14]. The pulse width of 0.4 ns was chosen. The interval between consecutive transmitted symbols was selected to be 40 ns. This interval is also equal to the 90% energy window length of the IRs. The bit rate of each user therefore is 25 Mbps. Perfect synchronization and no ISI effect are assumed in the simulation. The received signal of the th user has the form described in (3). The received signal is sampled at its peak. Then the sampled signal is detected based on the ideal decision threshold

Fig. 6. SIR with different settings and transmission schemes. (a) SIR of aligned time-reversed IRs, simple TR. (b)SIR of shifted and time-reversed IRs, the peaks are shifted by 2 ns relative to each other ( l , TR ns), TR&S.

1 =5 1

=2

we consider the delay time of 2.5 ms. Meanwhile the standard deviation value of the SIR is almost the same for both the perfect and outdated cases. Therefore, in the following analysis we use the instantaneous TR lter in the evaluation process and the results thereof could be considered as the upper bound of the TR performance. D. BER Simulation and System Throughput Using the measured channel IRs we simulated the average BER of the multiple users in a TR-UWB MISO system. At each measured location, the received signal power at the receiving antenna was calculated as the average power of the signal at its peak (12)

(14)

symbols were For each measured IR, some simulated in order to have enough data for statistical analysis. For each SNR value, the average BER is calculated as

(15)

3222

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

Fig. 8. Illustrations of the classical TH-UWB transmission scheme (top) and the TR-UWB transmission scheme TR&S-UWB (bottom).

Mbps, users with the total maximum bit rate of Fig. 8 top. The time slot for each particular user is assigned randomly from frame to frame depending on the TH-code. The total maximum bit rate of the system (all users) in a TH-UWB scheme is

(16)

Fig. 7. BER with different transmission schemes. (a) Aligned time-reversed IRs, simple TR. (b) Shifted and time-reversed IRs, TR&S.

where in this case is the number of simultaneous users. The results are shown in Fig. 7. The theoretical BER of the BPAM scheme in additive white Gaussian noise (AWGN) channel is plotted in the same gure for reference. When the transmitted time-reversed IRs are aligned in time at their peaks, the performance decreases signicantly for a large number of concurrent users. For more than 3 users and at a SNR value of 15 dB, the , Fig. 7(a). For the proBER already reaches a BER oor of posed TR&S scheme, the BER does increase with the number of users, but at a lower rate and the detected BER oor was below . At a SNR of 15 dB, a BER in the order of can be obtained for 5 simultaneous users, Fig. 7(b). As for the system capacity, we compare the total system bit rate for a multiuser system of the classical time hopping UWB (TH-UWB) transmission scheme [1], the TR-UWB and our proposed TR&S-UWB schemes. s For the TH-UWB scheme, with a frame interval ns, the system can serve 5 and time slot chip duration

where is the number of served users and is the bit rate of one user. Using the simple TR-UWB scheme we can pack more users into a single time slot. The signals of 5 (or more) users can be packed into the same time slot at the same epoch and therefore be overlapped in time, Fig. 2 (left). However, the performance in terms of the BER decreases fast with the number of users if the multiuser interference cannot be suppressed, Fig. 7(a). The maximum bit rate of the system (all users) in the simple TR-UWB scheme is (17) The number of simultaneous users is selected so that the requirement on the transmission quality of the system in terms of the BER is met. For example, if the target BER at 20 dB SNR then the allowable number of users is 4 and 3 for veris tical and horizontal polarization receiving antenna respectively, Fig. 7(a). In Fig. 8 (bottom), the signal assignments of the 5 users in the TR&S-UWB scheme are illustrated. For clarity, only the rst tap of the time-reversed IRs of the transmitted symbol is shown. With this TR&S-UWB scheme, it is also possible to pack 5 users in one time slot . However, because the interference is further suppressed, better BER performance can be achieved, Fig. 7(b).

NGUYEN et al.: A TIME REVERSAL TRANSMISSION APPROACH

3223

The maximum number of supported users in our proposed scheme (TR&S) can be roughly calculated as (18) where denotes the oor operator. The total bit rate of all users therefore is in a TR&S-UWB system consisting of (19) Roughly speaking, without taking the system performance in terms of the BER into account, using the proposed TR&S-UWB scheme we can increase the total system bit rate by a factor of as compared with the classical TH-UWB scheme. A combination of the TR&S-UWB scheme and the TH-UWB scheme is also a potential option. In this TR&S-TH-UWB scheme the time slot of each group consisting of users is assigned randomly depending on the TH-code. The advantage of this scheme is that the system can serve times more users. However, the system maximum total bit rate is similar to that of the TR&S-UWB scheme. V. CONCLUSION AND REMARKS In this paper we have proposed and evaluated the TR technique in UWB communications. We considered a particular multiuser TR-UWB system consisting of 4 transmitting antennas and 5 concurrent users each of them equipped with a single antenna (with either vertically or horizontally polarized elements). It was illustrated that the median SIR value observed at each user/location can be as large as 18 dB. Using the TR technique, spatial and temporal focusing is achieved, the receivers architecture becomes very simple and the required transmit power can be lowered. Using the proposed shifted time-reversed transmission scheme (TR&S), the simulation results showed that under certain assumptions, it is possible to communicate with 5 users simultaneously with an uncoded when the average SNR at the receiver is BER as low as 15 dB. With a typical loopback delay time of 2.5 ms, the effect of outdated channel information on the TR performance was shown to be negligible, namely 2 dB reduction in the mean SIR. For the same total throughput of 125 Mbps (5 25 Mbps), the BER versus the SNR performance of the proposed TR&S scheme is similar to the single user case with an All-Rake receiver [14], in an IEEE 802.15.3a multipath UWB channel [15]. In [8] similar performance is obtained for the TR SISO single user case in the measured multipath channel. These comparisons highlight the strong potential of TR schemes in MIMO/ MISO-UWB systems. Evaluating the performance degradation of the TR in UWB system when the imperfection in channel estimation is taken into account is one of the interesting problems for future work. Besides, using distributed instead of centralized antennas system at the AP in order to capture most of the waves emitted by the source/receiver so as to increase the spatial focusing could be an interesting investigation topic.

ACKNOWLEDGMENT The authors would like to thank the anonymous reviewers for their valuable comments and suggestions. H. T. Nguyen would also would like to thank his supervisors J. B. Andersen and G. F. Pedersen for their encouragement and fruitful discussions. REFERENCES
[1] M. Z. Win and R. A. Scholtz, Impulse radio: How it works, IEEE Commun. Lett., vol. 2, no. 2, pp. 3638, Feb. 1998. [2] M. Z. Win and R. A. Scholtz, Characterization of ultra-wide bandwidth wireless indoor channels: A communication-theoretic view, IEEE J. Select. Areas Commun. (JSAC), vol. 20, no. 9, pp. 16131627, Dec. 2002. [3] M. Fink, Time-reversed acoustic, Scientic Amer., pp. 6773, Nov. 1999. [4] A. Derode, A. Tourin, J. D. Rosny, M. Tanter, S. Yon, and M. Fink, Taking advantage of multiple scattering to communicate with time-reversal antennas, Phys. Rev. Lett., vol. 90, no. 1, pp. 014301-1014301-4, 2003. [5] P. Kyritsi, G. Papanicolaou, P. Eggers, and A. Oprea, MISO time reversal and delay-spread compression for FWA channels at 5 GHz, IEEE Antennas Wireless Propag. Lett., vol. 3, no. 6, pp. 9699, 2004. [6] S. M. Emami, J. Hansen, A. D. Kim, G. Papanicolaou, A. J. Paulraj, D. Cheung, and C. Prettie, Predicted time reversal performance in wireless communications using channel measurements, IEEE Communic. Lett., 2005, accepted for publication. [7] H. T. Nguyen, J. B. Andersen, and G. F. Pedersen, The potential use of time reversal technique in multiple elements antenna systems, IEEE Commun. Lett., vol. 9, no. 1, pp. 4042, Jan. 2005. [8] T. Strohmer, M. Emami, J. Hansen, G. Papanicolaou, and P. J. Arogyaswami, Application of time-reversal with MMSE equalizer to UWB communications, in Proc. IEEE Global Telecommunications Conf. (GlobeCom), Dec. 2004, vol. 5, pp. 31233127. [9] B. E. Henty and D. D. Stancil, Multipath-enabled super-resolution for rf and microwave communication using phase-conjugate arrays, Phys. Rev. Lett., pp. 243904-1243904-4, Dec. 10, 2004. [10] K. Usuda, H. Zhang, and M. Nakagawa, Pre-rake performance for pulse based UWB system in a standardized UWB short-range channel, in Proc. IEEE Wireless Communications and Networking Conf., Mar. 2004, vol. 2, pp. 920925. [11] S. Imada and T. Ohtsuki, Pre-Rake diversity combining for UWB systems in IEEE 802.15 UWB multipath channel, in IEEE Joint Conference on Ultrawideband Systems and Technologies, May 2004, pp. 236240. [12] My Personal Adaptive Global Net (MAGNET), IST 2004-507102, Oct. 2004, Deliverable D3.1.2a: PAN Radio Channel Characterisation (Part 1). [13] My Personal Adaptive Global Net (MAGNET), IST 2004-507102, Oct. 2004 [Online]. Available: www.ist-magnet.org, Deliverable D3.2.2a: Candidate Air-Interfaces and Enhancements [14] I. Oppermann, M. Hamalainen, and J. Linatti, UWB Theory and Applications. New York: Wiley, 2004, ISBN 0-470-86917-8. [15] J. Foerster, 802.15.3a Channel Modelling Sub-committee - Report Final Nov. 2002, IEEE P802.15 Working Group for WPAN.

Hung Tuan Nguyen (S03) received the B.Sc.E.E. degree in electronic engineering from Hanoi University of Technology, Vietnam, in 1999, the M.Sc.E.E. degree in telecommunication from the Technical University of Denmark, Lyngby, in 2002, and the Ph.D. degree from Aalborg University, Aalborg, Denmark, in 2005. He is now working as a Researcher at the Department of Communication Technology, Aalborg University. His research interest lies in MIMO radio channel modelling and characterizations, transmission techniques for multiple antennas system and studying the applicability of multiple antenna system in a small terminal.

3224

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

Istvn Zsolt Kovcs (M01) received the B.Sc. degree from Politehnica Technical University of Timisoara, Romania, in 1989, the M.Sc.E.E. degree from The Franco-Polish School of New Information and Communication Technologies/Ecole Nationale Suprieure des Tlcommunications de Bretagne, Poland/France, in 1996, and the Ph.D.E.E.degree in wireless communications from Aalborg University, Aalborg, Denmark in 2002. From 2002 to August 2005, he held the position of Assistant Research Professor with the Antennas and Propagation Division, Department of Communication Technology, Center For TeleInFrastruktur, Aalborg University. His research interests were in the eld of radio channel propagation measurements and modelling, with major focus on short-range ultrawideband (UWB) radio channel and ultrawideband antenna investigations. He was actively involved in the European IST PACWOMAN and IST MAGNET projects and participated in several industrial projects with partners such as TeleDanmark, Motorola, IOSpan, and ArrayComm. He has made a number of paper contributions and has contributed to three book chapters on UWB propagation topics. Currently, he is a Wireless Networks Specialist at Network Systems Research/Nokia Networks, Aalborg, conducting research in LTE UTRAN.

Patrick Claus Friedrich Eggers (M91) received the M.Sc.E.E. and Ph.D. degrees. Previously, he was an Associate Professor and had been Project Leader of the Propagation Group of CPK. He is now the Research Coordinator of the Antennas and Propagation Division, Department of Communication Technology (KOM), Aalborg University (AAU). He is a member of the Technical Research Council of the Center for TeleInFrastruktur (CTIF) at AAU, and has been Project and Work Package Manager in several European research projects (TSUNAMI, CELLO, etc.) and in industrial projects with partners as Nokia, Ericsson, Motorola, IOSpan, ArrayComm, Avendo Wireless, Samsung, etc., as representative of CTIF, KOM (and previously the center for PersonKommunikation-CPK). He is the author of over 40 papers, as well as section author and chapter editor in different COST nal reports (COST207, 231, 259) and books on ultrawideband and cooperative communication. He is the initiator and coordinator of an internationally targeted M.Sc.E.E. program in mobile communications taught in English at AAU, as well as Designer and Coordinator of the newly started program in Software Dened Radio.