The Vault

Shaping Spectral Leakage
Research Paper / Feb 2014

38 ||| 1556-6072/13/$31.00©2013ieee ieee vehicular technology magazine | september 2013 Digital Object Identifier 10.1109/MVT.2013.2269178 Date of publication: 25 July 2013 In this article, we propose a transceiver architecture based on orthogonal frequency-division multiplexing (OFDM) with transmit and receive windowing to reduce spec-tral leakage and reject adjacent channel interference (ACI). The transceiver enables the utilization of unused subbands in a fragmented spectrum with low complexity, which may be used in cognitive radio systems. We first show a multicarrier modulation (MCM) framework and the implementation of transmit and receive windowing. The intersymbol interference (ISI) resulting from receive windowing with a multipath chan- nel is illustrated, and the performance of the presented transceiver architecture is eval- uated and compared with several other MCM techniques. The results show that windowing offers a simple transceiver architecture with acceptable performance and is well suited for cognitive radio scenarios with noncontiguous spectrum fragments. Its performance can be further improved by using more advanced receivers. A Novel Low-Complexity Transceiver Architecture for Cognitive Radio erdem bala, Jialing li, and rui yang © photo f/x2 Image lIcensed by Ingram publIshIngShaping Spectral Leakage september 2013 | ieee vehicular technology magazine ||| 39 MCM is based on the idea of splitting a high-rate wide- band signal into lower-rate signals, where each signal occupies a narrower band called the subchannel [1], [2]. OFDM has proved itself as one of the most popular MCM techniques and is currently used in many wireless commu- nication systems such as the Third Generation Partnership Project (3GPP) long-term evolution (LTE) [3] and 802.11 [4]. In OFDM, a cyclic prefix (CP) is appended to the begin- ning of the data block before transmission. The CP, which is discarded at the receiver, provides several advantages, such as immunity against multipath and the possibility of using simple one-tap equalization for each subcarrier. In addition, OFDM enables efficient use of the available bandwidth through overlapping subcarriers. On the other hand, it has several disadvantages, such as spectral leak- age due to large side lobes of the subcarrier frequency response and high peak-to-average power ratio (PAPR). The demand for higher data rates has been increasing significantly. Several techniques have been studied and pro- posed to meet this demand. Among these, overlaying small cells over macrocells to allow spectral reuse, opening new bands to wireless communication, and using the available bandwidth more efficiently by spectrum sharing via cog- nitive radio are some of the notable techniques [5]. Since wireless systems are evolving toward a network of net- works architecture where many networks are expected to share the same spectrum, spectrally agile waveforms with small out-of-band emission (OOBE) are very important. To that end, the ACI created by the spectral leakage of OFDM makes this waveform unsuitable for these networks. As an alternative to OFDM, filter bank multicarrier (FBMC) schemes, specifically OFDM-offset QAM (OFDM- OQAM), have recently received great interest [6], [7]. OFDM-OQAM is another MCM technique where data on each subcarrier are shaped with an appropriately designed pulse so that side lobes are significantly reduced. A real data symbol is transmitted in each subchannel and on each OFDM-OQAM symbol, and consecutive OFDM-OQAM symbols are staggered. Adjacent subchannels overlap to maximize the spectral efficiency, creating intercarrier inter- ference (ICI); and several consecutive OFDM-OQAM sym- bols interfere with each other because of the long pulse, creating ISI. In a distortion-free channel, orthogonality can be achieved with a proper transceiver architecture, which can be efficiently implemented with polyphase filters. Al- though OFDM-OQAM offers significantly less spectral leak- age, its implementation in practical systems poses several challenges because of its complexity, latency, sensitivity to timing errors, and complex channel estimation and equal- ization algorithms in doubly dispersive channels [8]. There- fore, it is desirable to design an OFDM-like but spectrally contained waveform with improved OOBE characteristics. One way to improve the spectral containment of OFDM is filtering the OFDM-modulated signal, known as the filtered OFDM (F-OFDM). However, in a fragmented spectrum where available subbands are not contiguous, filtering becomes challenging since a separate filter needs to be designed and used for each subband [11]. More- over, filtering at the sampling rate results in high complex- ity. Another method is to use pulse shaping where the rectangular pulse shape used in conventional OFDM is smoothed to prevent sharp transitions between consecu- tive OFDM symbols, resulting in lower side lobes [9]–[11]. In an OFDM-based system, the intended receiver gets a com- pound of the desired and interfering signal and removes the samples corresponding to the CP as part of the receiver opera- tion. The removal of samples creates discontinuity in the sig- nal, resulting in an increase of the OOBE of the interfering signal even if it had very low OOBE at the transmission point. There- fore, a mechanism is required at the receiver to reject the ACI before the CP is removed. If filtering is used, the received signal should be filtered for individual subbands. Similar to transmit- ter filtering, receiver filtering imposes challenges in dynamical- ly available fragmented spectrum. An alternative method is to use a nonrectangular window at the receiver, which provides reduced side lobes while preserving orthogonality. Receive windowing has been used to reduce the impact of ICI due to carrier frequency offset or Doppler [13]–[16] and to suppress radio frequency interference (RFI) in discrete multitone (DMT) systems [17], [18]. In this article, we propose a transceiver architecture for OFDM-based systems with pulse shaping to reduce spectral leakage and receive windowing to reject ACI. The transceiver is shown to provide a low-complexity and effective way of using noncontiguous spectrum frag- ments in cognitive radio systems. Starting from an MCM framework as in [9], we first show that pulse shaping can be efficiently implemented with a specific type of windowing. Similarly, an efficient imple- mentation of receive windowing is presented. The resultant architecture, called windowed OFDM (W-OFDM), brings limited changes to a conventional OFDM transceiver and can be easily implemented. The OFDM transmitter and re- ceiver architecture implementations with windowing were previously presented in [9] through a filterbank approach. In most of the previous work, ISI due to receive window- ing in multipath channels is not considered. In this article, we also characterize the contribution of ISI due to receive windowing and evaluate its effect. The bit error rate (BER) performance of the presented transceiver architecture is evaluated and compared with other MCM techniques. In addition, a comparison of complexity and latency of differ- ent MCM techniques is presented. WindoWing offers a simple transceiver architecture With acceptable performance and is Well suited for cognitive radio scenarios With noncontiguous spectrum fragments. 40 ||| ieee vehicular technology magazine | september 2013 System Model In the following, k denotes the subcarrier index, n denotes the time sample index, and , denotes the symbol index. The input data sequence to be transmitted on the kth subcarrier and ,th symbol is denoted as [ ]Sk , . The total number of subcarriers is N . Figure 1 shows a discrete-time presentation of a block diagram for a general MCM scheme applicable to FBMC and OFDM. The data on each subcarrier are convolved by a filter [ ] [ ]p n p n ek j k fn2= Tr , where [ ]p n is the prototype filter, fT is the subcarrier spacing, and .F k fk T= The signaling period, and transmit and receive pro- totype filters differ according to the specific MCM scheme. Using the introduced notations, the critically sampled discrete-time OFDM signal can then be written as [ ] [ ] ,x n S p n N e ( ) k N k T j k N n 0 1 2 1,,= - , 3 3 ,r m = - =- -/ / 6 @ (1) where N N NCPT = + with NCP being the number of sam- ples in CP duration. The rectangular pulse shape [ ]p n is defined as [ ] , , , , , , otherwise. p n n N N1 0 1 1 0 CP f f = =- -) (2) Transmit and Receive Windowing Transmit Windowing It is known that the frequency response of the rectangu- lar pulse used in OFDM has high side lobes contributing to the spectral leakage to the adjacent bands. By smooth- ing the edges of the rectangular pulse, the spectral leakage can be reduced. One method to achieve this is to introduce a CP and a cyclic suffix (CS) before and after each data block, respectively, and shape the aggregate symbol by multiplying it with a smooth function, such as a raised cosine function. The CP consists of the samples at the end of the data block, while the CS consists of the samples at the beginning of the data block. Since the CP is multiplied with a nonunity function, orthogonality will be in general lost in a multipath channel. To prevent this, sometimes an extended guard interval (EGI) is added, which is effectively equivalent to lengthening the CP, and the original CP samples are kept outside the roll-of part of the windowing function. A sample windowing operation is illustrated in Figure 2(a) for a real signal. The overhead created by the CS results in spectral loss; one method to reduce this loss is to allow overlapping of the CP and CS of consecutive symbols, as shown in Figure 2(b). The implementation of transmit windowing can be de- rived from (1). Let us assume that the pulse shape [ ]p n is rectangular in the middle and smoothly converges to zero on the two sides, as illustrated in Figure 2(a). Such a pulse shape ensures smooth continuity between adjacent symbols, resulting in lower OOBE. One such common pulse shape is given in (3) in which the roll-off portions are of a raised cosine shape [ ] . , , . , . cos cos p n N n n N N n N N n N N n N 0 5 1 1 0 1 0 5 1 1 1 T T T T T T T T 1 1 # # # # r b b b r b b = + + + - + - `c c ^ j m m h Z [ \ ] ]] ] ]] ' ' _ ` a b bb b bb 1 1 (3) In (3), 0 1# #b is the roll-off factor and controls the length of the roll-off portion of [ ]p n , i.e., ( )N NCPb + . Let us define in (1) ,n iN mT= + where , ,m N0 1Tf= - , and i is an integer. Since for the ith block, only two symbols overlap because of the pulse shape, we keep the terms S0[l] S1[l] SN−1[l] S0[l] S1[l] SN−1[l] N p[n]ej2piF0n p[n]ej2piF1n p[n]ej2piFN−1n w[n]ej2piF0n w[n]ej2piF1n w[n]ej2pi N N + Channel N N N • • • • • • FN−1n pipi pipi pipi pipi pipi figure 1 an mcm transceiver block diagram. the overhead created by the cs results in spectral loss; one method to reduce this loss is to alloW overlapping of the cp and cs of consecutive symbols september 2013 | ieee vehicular technology magazine ||| 41 corresponding to i, = and i 1, = - in the summation over , . Then, we have [ ] [ ] [ ] [ ] [ ] . x iN m p N m S i e p m S i e 1 ( )T T k k M j k N N m k k M j k N m 0 1 2 1 0 1 2 1 T+ = + - + r r = - + = - / / (4) We can see from (4) that the pulse function and outputs of the inverse fast fourier transform (IFFT) for the ith and ( )i 1- th block are pointwise multiplied, i.e., windowed, and added. From (3), we see that the first term becomes nonze- ro only for , , , ,m N0 1 1Tf b= - which corresponds to the CS. Therefore, as illustrated in Figure 2(a) and (b), the win- dowed CS of the previous block is added to the windowed CP of the current block. The implementation is illustrated in Figure 2(c). The spectral agility of an MCM scheme can be seen by observing its OOBE. Figure 3 illustrates the power spec- tral densities of several MCM schemes in a noncontigu- ous spectrum. The two available subbands on the sides are used by an MCM transmitter while the subband in the middle is empty. For F-OFDM, a 35-tap root raised co- sine (RRC) filter with a roll-off factor of 0.05 covering the whole band is used. The length of the each roll-off por- tion of the W-OFDM is 128 samples. As shown in Figure 3, the spectral leakage of OFDM and F-OFDM into the middle subband is similar since one filter over the whole band cannot improve the OOBE in the middle of the available noncontiguous spectrum fragments. W-OFDM, on the other hand, is able to reduce the leakage with a single windowing operation since it is equivalent to filter- ing each subcarrier individually with a pulse that has a better frequency response than the rectangular pulse. Receive Windowing In an OFDM receiver, the CP is discarded, and the sam- ples corresponding to the data block are kept unchanged to go through the FFT. Therefore, this win- dow essentially consists of NCP zeros followed by N ones. Without proper ACI rejection, this window will cause the ACI to grow even if the signal from the adja- cent channel has low OOBE. Filtering is a common way to reject the ACI, but it introduces large computational complexity at the receiver. In addition, a separate filter needs to be used for each subband in a fragmented spectrum. Here we show that a window with a frequency response that has lower side lobes is better to reject the leakage from the adjacent channel. Let us define ,n iN m= + , , ,m N i0 1 1f= - as an integer, and [ ]w n as the receive window where some coefficients of [ ]w n can be zero. In the absence of transmit windowing, the esti- mate of the data symbol transmitted on subcarrier kt can be written as [ ] [ ] .S S w iN m e ( ) ( )k k i j k k N iN m m N k N 1 0 2 1 0 1 0 1 , ,= + r =- - + = - = - r t t6 @ /// (5) From (5), assuming that the window is of length N NCP+ and is applied to the CP and the data block, it can be shown that the condition to maintain the orthogonality is [ ] , , , .w iN m m N0 1constant i 1 0 f+ = = - =- / (6) The variable constant in (6) is the amplitude of the constant portion of the window function; the sum of the first NCP and last N samples of the window function should be equal to the constant portion of the window function. Note that, in a distortion-free channel, this op- eration does not change the data block because the sam- ples that are being added are the same and due to (6), the data block is effectively just scaled with a constant. The receive window used in conventional OFDM satisfies this condition. However, because of the discontinuity intro- duced by CP removal, its interference rejection capabili- ty is poor. The receive windowing operation is illustrated in Figure 4. Figure 4(a) shows an example of windowing operation for a real signal, and Figure 4(b) shows the implementation of receive windowing, which can be ob- tained from (5). Note that the window function in this figure satisfies the orthogonality condition and also has smooth transitions, providing better ACI rejection. First, the received OFDM symbol is pointwise multiplied by the receive window. Then, the samples corresponding to the CP are added to the tail of the symbol. The implementa- tion of receive windowing is illustrated in Figure 4(b). In general, a number of coefficients of the first NCP samples of the receive window may be zero; in this figure, the first a samples of the receive window are zeros. ISI Analysis Receive windowing may introduce ISI since part of the CP, which contains delayed copies of the transmitted signal, is multiplied with the window coefficients and added to the tail samples of the symbol. Figure 5 illus- trates the ISI contribution with an example. The channel in the figure consists of three paths; as- sume that the receiver is synchronized to the first path. The receiver receives the transmitted signal from the first path without delay and the two delayed copies from the other two paths. When the aggregate signal is win- dowed with the function given in the figure, the delayed W-ofdm, on the other hand, is able to reduce the leakage With a single WindoWing operation since it is equivalent to filtering each subcarrier individually. 42 ||| ieee vehicular technology magazine | september 2013 symbols arriving from the two paths will contribute to the ISI since they exist inside the CP. The amount of the ISI depends on the channel delay spread and the shape and length of the windowing function. There are several factors that are useful in lim- iting the contribution of the ISI. First, for many channels, the power of the paths decreases with increasing de- lay. Therefore, the ISI of the paths that are potentially most harmful is limited. Second, the roll-off portion of the window used to weigh the CP samples is an increasing function, limiting the contribution from the paths with higher power. Finally, the samples in the CP not used for win- dowing serves as a buffer against the multipath. In Figure 5, the second path does not contribute to the ISI since it is multiplied by zero-coeffi- cients of the window while the con- tribution of the third path is limited since it has lower power. There is a tradeoff between ACI rejection and ISI: a longer roll-off provides better ACI rejection; however, it also creates more ISI because of the increased ex- tent of the window. Since ISI depends on the previous symbol, it would be possible to can- cel this interference with a successive interference cancelation (SIC)-based receiver. As an alternative, the statis- tical information of the interference can be used by an MMSE receiver. Performance Evaluation In this section, we evaluate the BER performance, complexity, and laten- cy of W-OFDM, and compare it with OFDM, F-OFDM, and OFDM-OQAM. BER Performance The performance of the W-OFDM is evaluated and compared to the per- formance of OFDM, F-OFDM, and OFDM-OQAM when ACI exists. For the evaluation, a link-level simula- tion tool compliant with the 3GPP Release 8 LTE specifications is used. After QAM modulation, the transmit signal is generated by using one of the MCM schemes. The generated (b) CP CP Symbol Symbol CS CS (n−1)th OFDM Symbol After Windowing nth OFDM Symbol After Windowing nth Transmitted Symbol + S0[l] S1[l] SN−1[l] IFFT p[0] p[NCP−1] p[NCP] p[N−1] p[NT −1] p[NT] p[N] z −1 z −1 P/S x[n] + + (c) 20 10 0 −10 −10 −20 −20 1.5 1 0.5 20 10 0 0 0 10 20 30 40 50 60 70 80 0 10 20 30 Attach CP Attach CS 40 Tx Window Function The OFDM Symbol with CP and CS After Windowing 50 60 70 80 0 10 20 30 40 (a) 50 60 70 80 OFDM Symbol with CP (Red) and CS (Green) Attached p[βNT −1] p[NCP + βNT −1] p[N + βNT −1] p[(1 + β)NT −1] p[NCP + βNT] figure 2 an illustration of transmitter windowing and implementation. september 2013 | ieee vehicular technology magazine ||| 43 5 0 −20 −40 −60 −80 −100 −120 −4 −3 −2 −1 PS D (dB ) 0 Frequency (MHz) 1 2 3 4 OFDM F-OFDM OFDM−OQAM W-OFDM figure 3 the power spectral densities of several mcm schemes in fragmented spectrum. signal goes through a frequency- selective fading chan- nel. At the receiver side, the signal is first processed by the appropriate multicarrier demodulation, and then data symbols are demodulated. A spectral hole scenario similar to the one depicted in Figure 3 is created. The bandwidth of the whole chan- nel is set to 10 MHz and divided into 50 resource blocks (RBs), where one RB consists of 12 subcarriers. RBs 13– 36 are available and are used by the desired transmit- ter, while the remaining RBs are used by the interfering transmitter. There is a guard band of 1 RB on each side of the available band. In this setup, the interference created by the OOBE of the interferer leaks to the desired band from both sides. We assume that a 35-tap RRC filter cov- ering the whole channel is used for both transmit and receive filtering for the F-OFDM. The ACI is generated using the same waveform as the desired signal. The frequency offset between the desired and interfering signal is set to 0.5 fT , where fT is the subcarrier spacing and is equal to 15 kHz. The power difference between the ACI and the desired signal is a parameter denoted as PT and is set to 0 or 10 dB. An FFT size of 1,024 is used. The channel is mod- eled as the extended vehicular channel (EVA) with 5-Hz maximum Doppler [19]. A one-tap frequency do- main equalizer is used at the receiver, and ideal chan- nel estimation is assumed. The modulation scheme used is 16 QAM. Three types of windowing techniques are evaluated. In the first type (W-OFDM), both transmit and receive windowing are applied, and an EGI of 32 samples is added at the transmitter, i.e., effectively, a longer CP is used, resulting in a slight spectral efficiency loss. In the second type (W-OFDM/CP), the EGI is not added, and the CP is used for windowing purposes. Finally, in the third type, only receive (Rx) windowing is used (W- OFDM/Rx). The number of samples used for the roll-off portion of the windows is 32 and 64 for the transmit and receive windowing, respectively. The windowing function given in (3) is used for transmit windowing. For receive windowing, a similar function with length N NCP+ is used. Figure 6 compares the BER performance of the vari- ous MCM schemes when ACI dB.P 10T = We can see from this figure that OFDM performs the worst because of the significant leakage from the adjacent band. The performance of the F-OFDM is similar to OFDM since F-OFDM cannot use the spectral hole because of the fil- ter covering the whole 10 MHz band. The performance of the OFDM-OQAM is significantly better than both OFDM and F-OFDM because of its spectral agility and low OOBE characteristics. Since OFDM-OQAM does not have a CP, a one-tap equalizer is not optimal, and its performance may be further improved by using more complex equal- izers [20]. All three types of windowing techniques outperform OFDM, F-OFDM, and OFDM-OQAM. Out of the three W- OFDM types, W-OFDM/CP is the worst since the CP is used for windowing purposes, which degrades the effec- tiveness of the CP against multipath. In this case, one-tap equalizer is not optimal any more. W-OFDM with EGI is the best. A slightly longer EGI may be used to further re- duce the OOBE at the expense of a reduced spectral effi- ciency. Finally, the performance of W-OFDM/Rx illustrates the effectiveness of receive windowing for interference rejection. Although receive windowing introduces ISI, the benefit of rejecting ACI significantly outweighs the loss introduced by the additional ISI. Figure 7 compares the BER performance of the MCM schemes when ACI dB.P 0T = The observations are similar to those made for Figure 6 except that the per- formance of OFDM and F-OFDM is better because of the reduced ACI power. OFDM-OQAM and W-OFDM still perform better than OFDM and F-OFDM, but the perfor- mance gap is smaller. Note that all three types of win- dowing techniques outperform OFDM, F-OFDM, and OFDM-OQAM. Complexity and Latency Analysis In this section, we compare the complexity and latency of several MCM schemes. Complexity is evaluated in terms of number of real multiplications, but multiplica- tions with 1! and j! are not included since they are merely flips of sign and/or flips of real and imaginary WindoWing is a viable technique for aci rejection as Well as oobe reduction While preserving very loW complexity and latency. 44 ||| ieee vehicular technology magazine | september 2013 parts. A single MCM transceiver transmitting over the whole band with the parameters given in the previous section is assumed. For the complexity analysis of OFDM-OQAM, the implementation in [21] is assumed. For F-OFDM, it is assumed that the transmit filtering and addition of the CP could be combined such that the fil- tering is only done once for the CP samples [22]. For all schemes, the complexity of one-tap equalizer is also counted. The complexities normalized by the complexity of OFDM are presented in Table 1. We observe that 20 10 0 −10 −10 −20 −20 1.5 1 0.5 20 10 0 −10 −20 20 10 0 0 0 10 20 30 Received OFDM Symbol Including CP (Red) Rx Window Function Add CP Back Received OFDM Symbol with CP After Windowing Received OFDM Symbol After Adding the Windowed CP Back 40 50 60 70 80 0 10 20 30 40 50 60 70 80 0 10 20 CP to be Thrown Away 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 (a) w[a] w[a−1] w[NCP−1] w[N+NCP−1] w[N−1] w[N+a−1] w[N+a] w[NCP]y[n] S/P (b) + + FFT To the Equalizer Discard the First a Samples ... figure 4 an illustration of the receive windowing and implementation. september 2013 | ieee vehicular technology magazine ||| 45 compared to OFDM, both F-OFDM and OFDM-OQAM have relatively high complexity, while W-OFDM keeps the complexity increase to a minimum. Note that these numbers may slightly change depending on the specific implementation; however, it is clear that the complexity of W-OFDM is very close to that of OFDM. Latency is another important fac- tor to be compared, and we consider the inherent latency introduced in the MCM’s structure from the input of a QAM symbol to the output of the estimate of the QAM symbol. In OFDM, latency comes from the parallel- to-serial and serial-to-parallel conversion pair and CP and is equal to N NCP+ . Filtering or windowing does not introduce any additional latency. The latency of OFDM- OQAM, on the other hand, is . K N1 5 o+^ h , where Ko is the overlapping factor and is usually set to 3 or 4 for good OOBE performance [23]. Therefore, the latency of OFDM-OQAM is several times larger than that of OFDM for practical values of NCP and Ko . Conclusion In this article, a transceiver architecture based on trans- mit and receive windowing to reduce spectral leakage and reject ACI has been presented. This architecture has low complexity and therefore can be used to utilize non- contiguous fragmented frequency bands. The perfor- mance of the transceiver has been evaluated and compared with other MCM schemes. The results illus- trate that windowing is a viable technique for ACI rejec- tion as well as OOBE reduction while preserving very low complexity and latency. The performance may fur- ther be improved by using optimized window functions and advanced receivers to mitigate the ISI. Author Information Erdem Bala (erdem.bala@interdigital.com) received his B.Sc. and M.Sc. degrees from Bogazici Univeristy, Istanbul, Turkey, and his Ph.D. degree from the University of Dela- ware, all in electrical engineering. He has been with InterDigital Communciations, New York, as a research engineer since 2007. He has worked on several projects, including the standardization of 3GPP LTE and LTE-A, advanced relaying schemes, dynamic spectrum access, and, more recently, next-generation air interface systems. His previous work experience includes posi- tions as a software design engineer at the Turkey and U.K. research and development labs of Nortel Net- works and an internship at Mitsubishi Research Labs, Massachusetts. Jialing Li received her bachelor’s degree in elec- tronic engineering from the University of Science and Technology of China in 2005, her master’s of science in electrical engineering from the Polytechnic University in 2008, and her Ph.D. degree in electrical engineering from Polytechnic Institute of New York University (for- merly Polytechnic University) in 2011. She is currently a senior  engineer at InterDigital Communications Inc. Po w er Channel Power Delay Profile Receive Window Function nth Symbol TCP Delayed Version of the (n−1)th Symbol with Lower Power (n−1)th Symbol Time figure 5 an illustration of isi contribution due to receive windowing. 10 10−1 10−2 10−3 OFDM OFDM-OQAM F-OFDM W-OFDM W-OFDM/CP W-OFDM/Rx BE R 10−4 −5 0 5 10 15 20 25 Eb/No (dB) 30 35 40 45 OFDM OFDM-OQAMMMM F-OFDM W-OFDM W-OFDM/CPPP W-OFDM/Rx// figure 6 a comparison of various mcm techniques with aci .dBP 10T = figure 7 a comparison of various mcm techniques with aci .dBP 0T = 10 10−1 10−2 10−3 OFDM OFDM-OQAM F-OFDM W-OFDM W-OFDM/CP W-OFDM/Rx BE R 10−5 10−4 −5 0 5 10 15 20 25 Eb/No (dB) 30 35 40 45 46 ||| ieee vehicular technology magazine | september 2013 Her current research interests are coordinated multi- point transmission/reception, interference mitigation in heterogeneous networks, advanced waveform design for spectrally agile and power efficient systems, and physi- cal layer design of future spectrally efficient systems. Rui Yang  received his M.S. and Ph.D. degrees in electrical engineering from the University of Maryland in 1987 and 1992, respectively. He has 14 years of ex- perience in the research and development of wireless communication systems. Since he joined InterDigital Communications in 2000, he has led several product development and research projects. He is currently a senior engineering manager at InterDigital Innovation Lab, leading a project on baseband and RF waveforms for future wireless communication systems. His inter- ests include digital signal processing and air interface design for wireless devices, and he holds more than ten patents in those areas. References [1] A. Sahin, I. Guvenc, and H. Arslan. (2012, Dec. 14). A survey on proto- type filter design for filter bank based multicarrier communications. [Online]. Available: http://arxiv.org/pdf/1212.3374.pdf [2] B. Farhang-Boroujeny, “OFDM versus filter bank multicarrier,” IEEE Signal Processing Mag., vol. 28, no. 3, pp. 92–112, May 2011. [3] 3GPP, “Physical channels and modulation,” TS36.211, V10.4.0, Dec. 2011. [4] Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Standard 802.11™-2012, 2012. [5] B. Wang and K. J. R. Liu, “Advances in cognitive radio networks: A survey,” IEEE J. Select. Topics Signal Processing, vol. 5, no. 1, pp. 5–23, Feb. 2011. [6] P. Siohan, C. Siclet, and N. Lacaille, “Analysis and design of OFDM/ OQAM systems based on filterbank theory, IEEE Trans. Signal Processing, vol. 50, no. 5, pp. 1170–1183, May 2002. [7] M. Bellanger, D. LeRuyet, D. Roviras, M. Terré, J. Nossek, L. Baltar, Q. Bai, D. Waldhauser, M. Renfors, T. Ihalainen, A. Viholainen, T. H. Stitz, J. Louveaux, A. Ikhlef, V. Ringset, H. Rustad, M. Najar, C. Bader, M. Payaro, D. Katselis, E. Kofidis, L. Merakos, A. Merentitis, N. Passas, A. Rontogiannis, S. Theodoridis, D. Triantafyllopoulou, D. Tsolkas, D. Xenakis, M. Tanda, T. Fusco, M. Huchard, J. Vandermot, A. Kuzminskiy, F. Schaich, P. Leclair, and A. Zhao. (2010, June). FBMC physical layer: A primer, physi- cal layer for dynamic spectrum access and cognitive radio (PHY- DYAS). [Online]. Available: www.ict-phydyas.org/teamspace/ internal-folder/FBMC-Primer_06-2010.pdf [8] J. Louveaux, M. Tanda, M. Renfors, L. Baltar, A. Ikhlef, T. Hidalgo-Stitz, M. Bellanger, and C. Bader. (2009, July). Optimization of transmitter and receiver, PHYDYAS report. [Online]. Available: www.ict-phydyas. org/delivrables/PHYDYAS-D3-2.pdf [9] Y.-P. Lin, C.-C. Li, and S.-M. Phoong, “A filterbank approach to window designs for multicarrier systems,” IEEE Circuits Syst. Mag., vol. 7, no. 1, pp. 19–30, 2007. [10] T. Weiss, J. Hillenbrand, A. Krohn, and F. Jondral, “Mutual interfer- ence in OFDM-based spectrum pooling systems,” in Proc. IEEE 59th Vehicular Technology Conf., 2004, VTC 2004-Spring, 2004, Vol. 4, pp. 1873–1877. [11] P. Sutton, B. Ozgul, I. Macaluso, and L. Doyle, “OFDM pulse-shaped waveforms for dynamic spectrum access networks,” in 2010 IEEE Symp. New Frontiers in Dynamic Spectrum, Apr. 2010, pp. 1–2, 6–9. [12] D. Noguet, M. Gautier, and V. Berg. (2011). Advances in opportu- nistic radio technologies for TVWS, EURASIP J. Wireless Commun. Networking, 170. [Online]. Available: http://jwcn.eurasipjournals. com/content/2011/1/170 [13] C. Muschallik, “Improving an OFDM reception using an adap- tive Nyquist windowing,” IEEE Trans. Consumer Electron., vol. 42, pp. 259–269, Aug. 1996. [14] J. Armstrong, “Analysis of new and existing methods of reducing intercarrier interference due to carrier frequency offset in OFDM,” IEEE Trans. Commun., vol. 47, pp. 365–369, Mar. 1999. [15] S. H. Muller-Weinfurtner, “Optimum Nyquist windowing in OFDM re- ceivers,” IEEE Trans. Commun., vol. 49, no. 3, pp. 417–420, Mar. 2001. [16] N. C. Beaulieu and P. Tan, “On the effects of receiver windowing on OFDM performance in the presence of carrier frequency offset,” IEEE Trans. Wireless Commun., vol. 6, no. 1, pp. 202–209, Jan. 2007. [17] S. Kapoor and S. Nedic, “Interference suppression in DMT receivers using windowing,” in Proc. IEEE Int. Conf. Communications, vol. 2, pp. 778–782, 2000. [18] A. J. Redfern, “Receiver window design for multicarrier communi- cation systems,” IEEE J. Select. Areas Commun., vol. 20, no. 5, pp. 1029–1036, June 2002. [19] V. Erceg, K. V. S. Hari, M. S. Smith, D. S. Baum, K. P. Sheikh, C. Tappenden, J. M. Costa, C. Bushue, A. Sarajedini, R. Schwartz, D. Branlund, T. Kaitz, and D. Trinkwon (2001, July). Channel models for fixed wireless applications, IEEE 802.16.3c-01/29r4. [Online]. Avail- able: www.ieee802.org/16/tg3/contrib/802163c-01_29r4.pdf [20] J. Louveaux, L. Baltar, D. Waldhauser, M. Renfors, M. Tanda, C. Bader, and E. Kofidis, “Equalization and demodulation in receiver (single antenna),” PHYDYAS Rep. D3.1, July 2008. [21] A. Viholainen, M. Bellanger, and M. Huchard, “Prototype filter and structure optimization,” PHYDYAS Rep. D5.1, Jan. 2009. [22] D. Noguet, M. Gautier, and V. Berg, “Advances in opportunistic radio technologies for TVWS,” EURASIP J. Wireless Commun. Networking, vol. 170, pp. 1–12, Nov. 2011. [23] J. Fang, Z. You, J. Li, and I.-T. Lu, “Comparisons of filter bank multi- carrier systems,” in Proc. IEEE LISAT 2013. Table 1 Normalized complexity of MCM schemes. MCM Normalized complexity oFDm 1.00 F-oFDm (tx and rx filtering) 9.87 F-oFDm (tx filtering only) 5.28 oFDm-oQam 4.10 W-oFDm 1.01