ULTRA-WIDEBAND (UWB) communication systems

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 9, SEPTEMBER 2007 1667 Narrowband Interference Avoidance in OFDM-Based UWB Communication Systems Dimitrie C. Popescu, Senior Member, IEEE, and Prasad Yaddanapudi, Student Member, IEEE Abstract In this letter we present a new method for mitigating narrowband interference in orthogonal frequency-division multiplexing-based ultra-wideband communication systems. The proposed narowband interference avoidance (NBIA) method performs spectral shaping of the transmitted signal using binary signature sequences with minimum total squared correlation (TSC) to avoid the narrowband interfering signal. We illustrate the proposed method with plots that show the spectrum of the transmitted signal with and without NBIA, and also present numerical results obtained from simulations showing improvement in the bit error rate (BER) performance of the system when NBIA is employed. Index Terms Interference avoidance, orthogonal frequencydivision multiplexing (OFDM), spectral shaping, ultra-wideband (UWB). I. INTRODUCTION ULTRA-WIDEBAND (UWB) communication systems have generated increasing interest among researchers lately because of their potential for providing high data rates, and their robustness in multipath fading environments. UWB systems require large bandwidths for transmission of UWB signals (bandwidth larger than 500 MHz, or 25% of the center frequency), and must be capable of operating in the presence of various interfering signals coming from the existing (narrowband) communication systems. Thus, one of the challenges in the design of UWB communication systems is mitigation of narrowband interference. In our paper, we consider the orthogonal frequencydivision multiplexing (OFDM)-based UWB system proposed by Gerakoulis and Salmi [1], which was dubbed interference suppressing OFDM (IS-OFDM) system and which provides good performance in the presence of narrowband interference, and present a new method for mitigating narrowband interference in this system that further improves performance. The proposed NBIA method performs spectral shaping [2] in an OFDM framework, using binary spreading sequences with minimum TSC [3] [5] to avoid the spectrum of the narrowband interfering signal. We note that the concept of spectral shaping was introduced in the general context of UWB communication sys- Paper approved by G. Li, the Editor for Wireless Communication Theory of the IEEE Communications Society. Manuscript received August 8, 2005; revised April 4, 2006 and November 8, 2006. This work was supported in part by the Texas Higher Education Coordinating Board (THECB) Advanced Technology Program (ATP) under Grant 000512-0261-2003. This paper was presented in part at the GLOBECOM 2005 Conference. D. C. Popescu is with the Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, VA 23529 USA (e-mail: dpopescu@odu.edu). P. Yaddanapudi is with the Department of Electrical and Computer Engineering, University of Texas, San Antonio, TX 78249-0669 USA. Digital Object Identifier 10.1109/TCOMM.2007.904370 tems in [2], and is based on a spread-time code-division multiple access (CDMA) scheme [6], in which a signal is multiplied by a spreading sequence in the frequency domain as opposed to conventional spread spectrum systems in which the signal is multiplied by the spreading sequence in time domain. The idea behind spectral shaping is to use a spreading sequence with zeros in the frequency band corresponding to the narrowband interfering signal, which shapes the spectrum of the transmitted UWB signal to avoid the narrowband interference. For the IS-OFDM UWB system considered in our paper, spectral shaping is achieved by notching out several carriers that are in the same band as the narrowband interfering signal. This is accomplished by using the binary sequences with minimum TSC proposed by Karystinos and Pados [3] [5], instead of the Hadamard sequences used in [1]. The paper is organized as follows. In Section II, we give a brief description of the IS-OFDM UWB system and of binary sequences with minimum TSC. In Section III, we discuss how these sequences can be used to perform spectral shaping for the IS-OFDM UWB system and state the NBIA procedure. We continue with the presentation of numerical results obtained from simulations in Section IV, and present final conclusions in Section V. II. SYSTEM DESCRIPTION In the IS-OFDM UWB system, the wide transmission bandwidth is divided into Ñ frequency bins (or carriers) that are grouped into L groups with M frequency bins in each group, such that the total number of carriers Ñ = L M. The input data stream with rate R enters a serial-to-parallel (S/P) converter that provides L data streams each with rate R/L. Each parallel stream of rate R/L corresponding to a given group of frequency bins enters a second S/P converter, which provides M parallel streams each with rate R/Ñ. The M parallel streams in each group are spread by orthogonal Hadamard sequences {w q } M 1 0 of length M that returns the rate to R/L, and is followed by S/P conversion back to M parallel streams each with rate R/Ñ. These are then combined in an interference suppressing scheme by adding the orthogonally modulated symbols in a given group l to form a new set of M parallel symbols, such that the power of each of the M symbols in the frame carried by the given group of frequency bins is distributed over all M bins in the group, while symbols are separated by orthogonal Hadamard sequences in order to be able to distinguish them at the receiver and enable easy demodulation. The transmitter diagram for the IS-OFDM UWB system is depicted schematically in Fig. 1, and we refer the reader to the paper by Gerakoulis and Salmi [1] for complete details on the IS-OFDM UWB system. 0090-6778/$25.00 2007 IEEE

1668 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 9, SEPTEMBER 2007 asymptotic efficiency (TAE) [5]. Minimizing the TSC metric was used to design new binary valued spreading sequences [3], [4] that were shown to be optimal or near optimal with respect to the other criteria also [5]. The algorithm proposed by Karystinos and Pados [3], [4] yields a matrix of dimensions K L, whose K columns form a set of binary valued spreading sequences of length L with minimum TSC. The algorithm works for both underloaded case when K L, and overloaded case when K>N, and is initialized with a Hadamard matrix of size N =4 (max{k, L} +1)/4, where x rounds real number x to the nearest integer smaller than x. Fig. 1. The transmitter diagram for the IS-OFDM UWB system. The orthogonal Hadamard sequences used for spreading of the data streams in the IS-OFDM UWB system as described earlier, are binary valued sequences that are also used in spread spectrum systems for transmitting data over a larger bandwidth than is necessary, or to provide multiple access in orthogonal CDMA systems in which multiple users access the same frequency band at the same time. We note that spread spectrum transmission of data can also be achieved by using nonorthogonal binary valued spreading sequences like for example pseudorandom noise (PN) sequences. However, when nonorthogonal binary valued sequences are used to provide multiple access in (nonorthogonal) CDMA systems, each user s signal will act as interference for the signals of the other users, and the mutual interference between any two users depends on the cross correlation of their corresponding spreading sequences. Traditionally, PN sequences were used in the CDMA systems due to their special correlation properties: high correlation of a given PN sequence with itself and low correlation of the given PN sequence with other PN sequences or with shifted versions of itself. These properties imply relatively simple and efficient receiver structures based on single-user-matched filtering that can be used in a wide variety of scenarios. Recently, new performance metrics have been introduced in the CDMA literature to evaluate the overall performance of CDMA systems, like sum capacity, TSC, maximum squared correlation (MSC), or total III. NARROWBAND INTERFERENCE AVOIDANCE PROCEDURE The NBIA procedure proposed in our paper for the IS- OFDM UWB system is based on spectral shaping that was introduced by Da Silva and Milstein in [2], as a method for mitigating narrowband interference in UWB communication systems. The basic idea behind spectral shaping is to multiply the transmitted signal in frequency domain by a spreading sequence with zeros in the frequency band corresponding to the narrowband interfering signal. This operation shapes the spectrum of the transmitted UWB signal to avoid the narrowband interference. For the IS-OFDM UWB system, spectral shaping can be achieved by avoiding carriers that are in the same frequency interval as the narrowband interfering signal. This can be accomplished by replacing the Hadamard sequences {w q } M 1 0 in frequency bin l where the narrowband interfering signal is located, by alternative sequences that have zeros in places corresponding to those particular carriers where the spectral gap in the transmitted signal is required. As mentioned in [2], this replacement has a twofold effect: it shapes the spectrum of the transmitted signal to avoid the narrowband interference, and at the same time, suppresses the narrowband interference at the receiver that uses the same spreading sequences in the demodulation process. However, replacing Hadamard sequences by alternative sequences spoils the orthogonality of symbols transmitted over the same carrier, and creates interference among them at the receiver that affects performance. The level of interference depends on the cross-correlation properties of the set of alternative sequences, and is usually quantified by the TSC. Among all binary sequences (including the various PN sequences like maximal length, Gold, etc.), those developed by Karystinos and Pados [3], [4] have minimum TSC and allow decoding of a given symbol under minimum total interference from the other transmitted symbols. In the NBIA procedure, we use these binary sequences with minimum TSC developed by Karystinos and Pados [3], [4] to perform spectral shaping of the transmitted UWB signal. We note that the interference introduced by replacing the orthogonal Hadamard sequences with alternative nonorthogonal binary sequences introduces an error floor in the BER performance of the IS-OFDM system even in the absence of the narrowband interfering signal, as can be seen in Fig. 2 that illustrates the BER performance of the IS-OFDM system when Hadamard sequences are used for spreading in all

POPESCU AND YADDANAPUDI: NARROWBAND INTERFERENCE AVOIDANCE 1669 Fig. 2. BER performance of the IS-OFDM UWB system in the absence of narrowband interference for several types of spreading sequences. the L groups of frequencies and when alternative sequences are used in one of the groups. 1 The NBIA procedure for spectral shaping, using binary spreading sequences with minimum TSC, requires knowledge of the narrowband interfering signal and/or of its position in the frequency domain. This knowledge can be obtained by using various techniques, for example, comparing the frequency domain of the signal at the receiver with an interference threshold derived using a spectral template of the desired signal [7] [9]. In this case, the carriers that are above the interference threshold are considered to be affected by the narrowband interference, and will be subject to the NBIA procedure. In order to present the NBIA procedure, we assume that narrowband interference has been detected in the band of frequencies [f 1,f 2 ]. Then, the bandwidth of the narrowband interfering signal is equal to F NBI = f 2 f 1 (1) and the signal is located in the IS-OFDM frequency bin number f 1 l = (2) F IS-OFDM where F IS-OFDM = F total /L, F total is the total bandwidth of the UWB signal, and x rounds real number x to the nearest integer larger than x. In terms of the subcarriers in bin l, the narrowband interfering signal spectrum starts at carrier number f1 (l 1) F IS-OFDM n s =. (3) f The total number of carriers that must be avoided is FNBI k f = +1 (4) f where f is the width of a single carrier. The narrowband interfering spectrum ends at carrier number n e = n s + k f (5) Using the parameters in (3) (5), the NBIA procedure can be formally stated as follows. 1) Start with an M M Hadamard matrix and use the algorithm of Karystinos and Pados in [3], [4] to construct an M ( M k f ) matrix S of binary sequences with minimum TSC. 2 2) Augment matrix S to an M M matrix S by placing k f columns with all elements equal to zero in those places corresponding to the k f subcarriers to be avoided. 3) Use columns of matrix S to replace the Hadamard sequences {w q } M 1 0 in frequency bin l. We note that orthogonal Hadamard sequences will be used in the rest of the IS-OFDM frequency groups. IV. SIMULATION RESULTS We considered an IS-OFDM UWB system with bandwidth F total = 1.25 GHz identical to that in [1], divided into Ñ = 512 frequency bins grouped into L =8groups, each with M =64 subcarriers. This implies that each frequency bin is f = 2.4414 MHz wide, and that the bandwidth of a group of frequency bins is approximately F IS-OFDM = 156.25 MHz. We assumed a narrowband interfering signal (jammer) with bandwidth F NBI = 5 MHz, located between frequencies f 1 = 500 MHz and f 2 = 505 MHz, which was generated similar to [1] by using a linear bandpass FIR filter with passband equal to 5 MHz driven by white Gaussian noise with unit variance at the input. From (2), the group number where the jammer is located is obtained as l =4. Using (3) (5), the subcarrier number where the jammer starts is n s =12and the number of subcarriers it overlays is k f =4. Thus, the jammer spectrum overlays over frequency bins 12 through 15 in the fourth IS-OFDM group. The total power of the jammer is selected relative to that of the UWB signal such that specific values of the jammer-to-signalratio (JSR) are obtained. We apply the NBIA procedure described in the Section III and construct the M ( M k f ) (that is, 64 60 for our numerical example) matrix S of binary spreading sequences using the algorithm in [3] and [4]. Next, we augment S by adding k f =4 column vectors with all elements zero after column 11 to obtain matrix S of dimension M M (that is 64 64 for our numerical example) whose columns will be used to replace the Hadamard sequences in the IS-OFDM group number n =4. To illustrate the spectral shaping effect of the NBIA procedure, we have plotted the power spectral density of the transmitted signal before and after the procedure was applied in Fig. 3. 1 The simulation setup used in obtaining Fig. 2 is described in detail in Section IV. 2 This corresponds to the overloaded case with K>Lin the notation of Karystinos and Pados [3], [4].

1670 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 9, SEPTEMBER 2007 Fig. 3. Power spectral density of the transmitted signal before and after the NBIA is performed. Fig. 4. BER performance of the IS-OFDM UWB system with AWGN and narrowband interferer, before and after NBIA is applied. The plot shows the spectral gap that occurs in the spectrum of the transmitted UWB signal when NBIA is employed, in the frequency band corresponding to the jammer. We have also performed a series of simulations to investigate the BER performance of the considered IS-OFDM UWB system with and without application of the NBIA procedure. We used the IS-OFDM UWB receiver presented by Gerakoulis and Salmi [1] and looked at the raw BER after demodulation/detection without considering any error correction techniques. We note that the use of error correction techniques will only improve the raw BER observed after demodulation/detection. In the first simulation experiment, we did not consider the jammer present and looked at the BER with AWGN only when Hadamard sequences are used in all groups of frequencies, as well as when PN sequences and binary sequences with minimum TSC are used in the fourth group of frequencies while Hadamard sequences are used in all other groups. The results of this experiment are shown in Fig. 2, which shows the error floor introduced by the use of nonorthogonal spreading sequences beyond which the BER cannot be decreased by increasing E b /N 0. We note that this floor does not depend on the presence of the narrowband interfering signal, and is minimum for binary sequences with minimum TSC. In the second experiment, we simulated the system in AWGN with the jammer active, for JSR values of 5, 10, and 30 db, and looked at the BER before and after application of the NBIA procedure. The results of this simulation are presented in Fig. 4. The dashed curves in Fig. 4 correspond to the BER before application of NBIA, and also show an error floor beyond which the BER will not be improved by increasing E b /N 0.Inthis case the error floor is due to the spectral leakage caused by the discontinuity in the discrete Fourier transform (DFT) blocks that occurs as a result of the stochastic nature of the narrowband jammer that does not have an integer number of cycles in the DFT block. This spectral leakage affects all the subcarriers in the transmitted signal not only those subcarriers colocated with the narrowband interference in the frequency domain. As can be seen from Fig. 4, the error floor introduced by the spectral leakage in AWGN is higher than that introduced by the use of binary sequences with minimum TSC seen in Fig. 2, and is especially significant at high JSRs. The solid curves in Fig. 4 correspond to the BER after application of NBIA, and show that the error floor displayed is smaller than when NBIA is not applied. We note that for low and average JSR (5 and 10 db), applying NBIA leads to an approximately one order of magnitude decrease in the BER floor, while for large JSR, (30 db), the BER improvement is lower. We also note that the BER floor after application of NBIA for large JSR values can be further decreased by compensating the effect of spectral leakage on the recovered symbols. This can be achieved by removing extra subcarriers adjacent to the jammer as in [8], or by using a windowing technique at the receiver [10]. Fig. 5 illustrates the reduction in error floor after NBIA is applied for high JSR values (20 and 30 db), when additional adjacent subcarriers in the UWB signal are nulled for spectral leakage compensation. In the third experiment, we looked at the BER performance in the presence of multipath and simulated the system using a multipath channel between transmitter and receiver as described in [11] for the same JSR values of 5, 10, and 30 db. Results of this experiment are presented in Fig. 6, and similar observations as in the previous case can be made that the use of the NBIA procedure improves the BER performance. To provide a better view of the performance improvement implied

POPESCU AND YADDANAPUDI: NARROWBAND INTERFERENCE AVOIDANCE 1671 Fig. 5. BER performance of the IS-OFDM system with AWGN and narrowband interferer for high JSR, with NBIA and spectral leakage compensation. Fig. 7. BER performance versus JSR for several E b /N 0 values for the IS-OFDM UWB system with multipath and narrowband interferer. Fig. 6. BER performance versus E b /N 0 for several JSR values for the IS- OFDM UWB system with multipath and narrowband interferer. by NBIA for multipath channel and narrowband jammer, we simulated the system also for varying values of JSR and fixed E b /N 0 =5, 10, 15, 20, 25, and 30 db, and the results of this simulation are shown in Fig. 7. We note that, for a given value of E b /N 0, the improvement in BER decreases with the increase in JSR, which may be due to the fact that, at higher JSRs, the spectral gap created in the transmitted signal may not be deep enough to cancel entirely the spectral peak of the jammer. In the fourth experiment, we looked at the BER performance of the system when more than one narrowband jammer is active. We considered two distinct cases: one in which we assumed that there are two narrowband jammers in the same IS-OFDM frequency bin and another one in which the two narrowband jammers are located in distinct IS-OFDM frequency bins. The results of this experiment are shown in Fig. 8. We note that when there are two narrowband jammers in distinct IS-OFDM, frequency bins applying the NBIA procedure yields over one order of magnitude decrease in the BER floor that becomes close to that of the system with only one narrowband jammer and NBIA. When the two narrowband jammers are in the same IS-OFDM frequency bin, the BER floor is still reduced, although the gain is less than in the previous case. This decrease in performance may be overcome by increasing the number of IS-OFDM frequency bins L, so that the narrowband jammers lay over different frequency bins. Finally, in the last experiment of this series, we compared the performance of the IS-OFDM system with and without the proposed NBIA procedure with that of a conventional OFDM-based UWB system with frequency domain interleaving followed by convolutional coding. For the conventional OFDM-based UWB system, we consider a system with the same total bandwidth of 1.25 GHz and the same number N = 512 of subcarriers. We employ frequency domain interleaving with a depth of two followed by a convolutional coding with rate R =1/2. Theresults of this experiment are shown in Fig. 9. We note that when the systems operate only in AWGN, at low JSR (5 db), the IS-OFDM system with NBIA has the best performance followed by the IS-OFDM system without NBIA, and both IS-OFDM systems outperform the conventional OFDM-based UWB system. At higher JSR (10 db and above), the conventional OFDM

1672 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 9, SEPTEMBER 2007 We conclude this section by noting that we have also looked at the sensitivity of the NBIA procedure, when the position of the jammer bandwidth is not identical to the bandwidth that the NBIA procedure avoids. For this, we simulated the system to determine the variation in BER performance improvement implied by the proposed NBIA procedure with various offsets of the actual position of the center frequency of the narrowband interfering signal from the value (f 1 + f 2 )/2 corresponding to the center of the frequency band avoided by the NBIA procedure. Our simulations have shown that when the bandwidth of the narrowband jammer overlaps over about 70% or more with the frequency band avoided by the NBIA procedure, the improvement in BER implied by NBIA is essentially not affected by the offset. Fig. 8. BER performance of the IS-OFDM system with more than one narrowband interferer. V. CONCLUSIONS In this paper, we presented a new procedure for avoiding narrowband interfering signals in OFDM-based UWB communication systems. The NBIA procedure performs spectral shaping [2] of the transmitted signal using the newly developed binary spreading sequences with minimum TSC [3] [5] to avoid those subcarriers located in the same frequency band as the narrowband interfering signal. The novelty of the proposed NBIA procedure consists in application of the spectral shaping concept in an OFDM framework that is different from [2], as well as in the new way the binary spreading sequences with minimum TSC are used, that is different from their original intended use of providing multiple access in CDMA systems. We note that application of the NBIA procedure in the IS-OFDM system trades off orthogonality of transmitted symbols in group l of frequency bins where the narrowband interference is present for improved performance in the presence of the narrowband interference. We illustrated the proposed NBIA procedure with an example that displays the spectrum of the transmitted signal with and without NBIA and shows the gap in the spectrum of the transmitted signal when NBIA is performed. This occurs in the frequency band of the narrowband interfering signal and implies that the transmitter avoids sending information in this band. We also presented numerical results obtained from simulations that show that the BER performance of the considered OFDM-based UWB system is improved when NBIA is employed, and that it is superior to that of conventional OFDM-based UWB system that uses frequency domain interleaving and convolutional coding. Fig. 9. BER performance for OFDM UWB system with frequency interleaving and convolutional coding, IS-OFDM UWB system, and IS-OFDM UWB system with NBIA. system has performance very similar to that of the IS-OFDM system without NBIA, while the IS-OFDM UWB system employing NBIA still continues to have the best performance. We also note from Fig. 9 that even in the case of a multipath channel, the IS-OFDM UWB system employing the NBIA procedure continues to display the best performance, although in this case, the differences were not so significant. ACKNOWLEDGMENT The authors are grateful to the anonymous reviewers for their constructive comments on the paper. REFERENCES [1] D. Gerakoulis and P. Salmi, An interference suppressing OFDM system for ultra wide bandwidth radio channels, in Proc. 2002 IEEE Conf. Ultra Wideband Syst. Technol., Baltimore, MD, pp. 259 264. [2] C. Da Silva and L. B. Milstein, Spectral-encoded UWB communication systems, in Proc. 2003 IEEE Conf. Ultra Wideband Syst. Technol., Reston, VA, pp. 96 100.

POPESCU AND YADDANAPUDI: NARROWBAND INTERFERENCE AVOIDANCE 1673 [3] G. N. Karystinos and D. A. Pados, Minimum total squared correlation design of DS-CDMA binary signature sets, in Proc. 2001 IEEE Global Telecommun. Conf., San Antonio, TX, vol. 2, pp. 801 805. [4] G. N. Karystinos and D. A. Pados, New bounds on the total squared correlation and optimum design of DS-CDMA binary signature sets, IEEE Trans. Commun., vol. 51, no. 1, pp. 48 51, Jan. 2003. [5] G. N. Karystinos and D. A. Pados, The maximum squared correlation, sum capacity, and total asymptotic efficiency of minimum total squared correlation binary signature sets, IEEE Trans. Inf. Theory,vol.51,no.1, pp. 351 354, Jan. 2005. [6] P. Crespo, M. Honig, and J. Salehi, Spread-time code-division multiple access, IEEE Trans. Commun., vol. 43, no. 6, pp. 2139 2148, Jun. 1995. [7] E. Darbyshire, Narrowband interference identification and rejection applied to baseband radio signals, in Proc. 1989 Inst. Electr. Eng. 5th Int. Conf. Radio Receiv. Assoc. Syst., Cambridge, U.K., pp. 217 221. [8] K. Shi, B. Kelleci, T. W. Fischer, Y. Zhou, E. Serpedin, and A. Karsilayan, On the design of robust multiband OFDM ultra-wideband receivers, presented at the 2005 Texas Wireless Symp., University of Texas at Austin, Austin, TX. [9] M. Sahin and H. Arslan, A narrowband interference identification approach for UWB systems, in Proc. 2005 IEEE Mil. Commun. Conf., Atlantic City, NJ, pp. 1404 1408. [10] A. J. Redfern, Receiver window design for multicarrier communication systems, IEEE J. Sel. Areas Commun., vol. 20, no. 5, pp. 1029 1036, Jun. 2002. [11] A. Batra, J. Balakrishnan, G. R. Aiello, J. R. Foerester, and A. Dabak, Design of a multiband OFDM system for realistic UWB channel environments, IEEE Trans. Microw. Theory Tech.,vol. 52,no.9,pp.2123 2138, Sep. 2004.