V3-219 Frequency Domain Channel Estimation and Detection for Impulse Radio Systems Andrea M. Tonello and Roberto Rinaldo Dipartimento di Ingegneria Elettrica Gestionale e Meccanica (DIEGM) - Università di Udine Via delle Scienze 208 - 33100 Udine - Italy phone: +39 0432 558 288 - fax: +39 0432 558 251 - e-mail: tonello@uniud.it Abstract—This paper deals with channel estimation, and detection in ultra wide band (UWB) bi-phase impulse modulated systems. We address the single user and the multiuser scenario assuming a direct sequence spreading code division multiple access (DS-CDMA) scheme. The channel estimation and detection approach is single user based, and operates in the frequency domain. In the presence of multiple access interference (MAI) the algorithm is appropriately modified to include the capability of canceling the interference through the exploitation of its frequency domain correlation. The approach can be extended to time-hopped impulse radio systems. Keywords—CDMA, Channel estimation, Frequency domain processing, Impulse modulation, Interference Cancellation, Multiuser Detection, Synchronization, Ultra wide band (UWB). I. INTRODUCTION 1 T HIS paper deals with synchronization, channel estimation, and detection in impulse radio systems. Several combinations of modulation, and user multiplexing schemes have been proposed for impulse radio communications [3]. The common attractive feature is the carrier-less baseband implementation that involves transmission of short duration pulses. This technology is commonly referred to as ultra wide band (UWB) because the pulses can occupy a very large bandwidth. Most of the work has focused so far on schemes that deploy time hopping spreading codes with pulse position modulation. Instead, in this paper we assume bi-phase pulse modulation (BPAM) in conjunction with direct sequence code division multiplexing of users (DS-CDMA) [3], [7]. Binary codewords are assigned to users, and modulate short duration pulses (monocycles). A user’s codeword spans a transmission frame. Frames are separated by a guard time to cope with the channel time dispersion. When the guard time is longer than the channel time dispersion, and only a single user accesses the medium, the optimal receiver comprises a matched filter followed by a symbol by symbol threshold detector [1]. The receiver filter has to be matched to the equivalent impulse response that comprises the user’s waveform, and the channel impulse response. Since UWB signals can occupy a large bandwidth, the channel is highly frequency selective, and the received signal exhibits a large number of multipath components. Potentially, high frequency diversity gains can be achieved. However, the optimal matched filter receiver has to accurately estimate the channel, and such an estimation can be particularly complex if performed in the time domain. It has been shown in [13] that channel estimation can be partitioned into a two step process if we model it as a tapped delay line. That is, we first determine the channel ray delays, and then we obtain an estimate of the ray amplitudes. Unfortunately, the ray search has a complexity that grows exponentially with their number. Further, false ray detection may occur in the absence of a priori knowledge about the true number of rays. The search can be partially simplified under the assumption of the channel to be separable [6], [13]. However, this assumption translates into deep performance losses in the non-rare event of clusters of non-separable rays. 1 Part of this work was supported by MIUR under project FIRB "Reconfigurable platforms for wideband wireless communications", prot. RBNE018RFY. When the common media is shared by multiple users, multiple access interference (MAI) may arise at the receiver side. In a DS- CDMA system, this is due to the deployment of non orthogonal codes, or to users that are time asynchronous, or to the presence of channel time dispersion. Assuming a single user detection approach the MAI translates into performance losses, such that some form of multiuser detection is advisable. Motivated by the above considerations, we propose a frequency domain approach to channel estimation, and detection [2], [12]. The approach is single user based. However, it can include the capability of rejecting the MAI interference. It has been derived from the observation that the optimal matched filter receiver can be equivalently implemented in the frequency domain. The approach comprises the following stages. First we acquire frame synchronization with the desired user. Second, we run a discrete Fourier transform (DFT) on the received frames. Third, we perform frequency domain channel estimation for the desired user via a recursive least squares (RLS) algorithm. Finally, detection is accomplished in the frequency domain using the estimated channel frequency response. Frame timing is crucial. In this paper we also address this problem and we describe a frame timing algorithm. In the presence of multiple access interference the algorithm is appropriately modified to include the capability of canceling the interference. Interference rejection is accomplished by observing that the MAI manifests itself with a frequency domain correlation that can be estimated and exploited by the detector. II. FREQUENCY DOMAIN PROCESSING In our system model (Fig. 1) we assume bi-phase pulse amplitude (BPAM) modulation [9] such that the signal transmitted by user u can be written as () ( ) u u u k f k s t bg t kT = − ∑ (1) where 1 u k b = ± denotes the information bit transmitted in the k-th frame, () u g t is the waveform used to convey information for user u, and f T is the bit period (frame duration). We further deploy direct sequence spreading to accommodate for multiplexing of users [3]. The user’s waveform (signature code) comprises the weighted repetition of 1 L ≥ narrow pulses (monocycles), i.e., 1 0 () ( ) L u u m M m g t cg t mT − = = − ∑ (2) DS-Spread User 0 Pulse Channel User 0 + k b DS-Spread User 1 Pulse Channel User 1 1 k b DS-Spread User N I Pulse Channel User N I I N k b Front-End Filter Sampler S/P M-Point FFT Frame Synchronization FD MF FD Channel MAI Correlation Estimation 0 k bc 0 T f T g T 1 k bc 1 k L bc − Fig. 1. Impulse modulated system with frequency domain (FD) processing, and frame structure. Proceedings of International Symposium on Wireless Personal Multimedia Communications 2004, WPMC 04 Abano Terme, Italy - September 12-15, 2004