NARROWBAND INTERFERENCE SUPPRESSION IN MULTI-BAND OFDM ULTRA WIDEBAND COMMUNICATION SYSTEMS: A MIXED-MODE APPROACH Burak Kelleci, Timothy W. Fischer, Kai Shi, Yi Zhou, Aydın ˙ Ilker Kars ¸ılayan and Erchin Serpedin Department of Electrical Engineering Texas A&M University, College Station, TX 77843-3128 serpedin@ece.tamu.edu ABSTRACT To avoid significant losses that might be induced by a narrow- band interference (NBI) on the performance of a multi-band (MB) OFDM ultra wideband (UWB) receiver, efficient NBI mitigation schemes are required. This paper proposes a novel mixed-mode NBI suppression scheme that relies on the cooperation between a digital NBI detector and an adaptive analog notch filter. Sim- ulation results show that the proposed mixed-mode suppression scheme improves the performance of the digital frequency exci- sion scheme in a MB-OFDM UWB receiver by as much as an equivalent signal-to-interference ratio gain of 9 dB. 1. INTRODUCTION Ultra-wideband (UWB) transceivers hold the promise to revolu- tionize the next generation of short-range wireless networks. For the UWB transceivers to coexist with nearby devices, it is neces- sary to design efficient UWB receivers whose operation is robust to narrowband interferences (NBI). References [1]–[3] outline the most relevant NBI mitigation schemes, which can be broadly clas- sified into two main groups. The first class of methods is moti- vated by the fact that the received signal consists of a correlated NBI and an uncorrelated spread spectrum signal. A pre-whitening filter [3]–[5] can be used to reduce the correlation of the received signal, and thus to suppress the NBI. The pre-whitening filter is found to improve the performance of spread spectrum receivers especially when the interference is strong [3]. However, the un- derlying assumption required by this class of methods can not be further extended to multi-band OFDM based UWB transceivers since the OFDM signal is correlated during one symbol period. When compared to the desired wideband signal, the interfer- ence occupies a much smaller frequency band and presents a higher power spectral density. Motivated by this fact, the second class of methods relies on transform domain filtering techniques to sup- press the NBI. By converting the received signal into the frequency domain via real-time transform devices [4], one can easily deter- mine the frequency location of NBI. The part of the signal cor- rupted by NBI is excised in the frequency domain and the remain- ing signal is re-transformed into the time domain. This method appears to be also fit for NBI suppression in OFDM receivers. In this paper, a mixed-mode (hybrid) interference mitigation scheme that consists of a digital NBI detector and an adaptive ana- log notch filter (AANF) is proposed. Our study shows that such This work was supported in part by the Texas Higher Education Coor- dinating Board and NSF. mixed schemes could be applied with success to combat strong NBI (i.e., interferences that give rise to signal-to-interference ra- tios (SIR) less than 0 dB) in multi-band (MB) OFDM UWB re- ceivers. It is also found that for combating mild interferences (i.e., those interferences for which SIR ≥ 0 dB), it is sufficient to con- sider a low-complexity digital frequency excision method. Our comprehensive simulations show that in the presence of strong NBI, a mixed mode NBI mitigation scheme provides a significant gain (an equivalent SIR gain of 9 dB) relative to a receiver that as- sumes only a digital frequency excision method. In the following sections, operation of the digital and analog suppression schemes are discussed, and the performance of the mixed-mode NBI miti- gation system is demonstrated. 2. DIGITAL NBI DETECTION AND MITIGATION The detection of NBI requires a fast Fourier transform (FFT) de- vice, which is already present in an OFDM receiver. Since the phase of NBI is unknown, the NBI has to be detected noncoher- ently based on the absolute values of the received signal samples. Without any loss of generality, the NBI is modeled as a complex sinusoid with unknown amplitude, frequency and phase [6]. For flat-fading channels the highest peak corresponds to the subcarrier affected by NBI. Therefore, by comparing the magnitude-squared of the FFT bins |X[k]| 2 with a threshold T0, one can easily find out not only the existence of NBI but also the subcarrier location of NBI. The performance of the NBI detector depends on the value of threshold: a too small T0 could lead to large false alarms, while a larger T0 could increase the probability of a miss. In general, the Neyman-Pearson detection framework could be used to opti- mize the value of threshold T0. In frequency-selective channels the OFDM signal presents a large dynamic range, and some signal peaks could take values close to the peak induced by NBI. This observation indicates that the detection of NBI becomes difficult in frequency-selective channels, especially when the SIR is large. Fortunately, in packet-oriented UWB systems, for most of the time the channel is clear of the transmitted signal. To avoid the effects introduced by the OFDM signal, we may detect the NBI during the time interval between the two packets. Since the subcarriers corrupted by NBI are not reliable any- more, they should not be used during the demodulation. The meth- od of erasing the corrupted subcarriers is referred to as frequency excision, and the lost data caused by excision can be recovered by exploiting channel decoding techniques. To evaluate the per- formance of the frequency excision method, we run simulations assuming an MB-OFDM receiver. It is well known that, due to 55 1-4244-0535-1/06/$20.00/©2006 IEEE