An Efficient Spectral Leakage Filtering for IEEE 802.11af in TV White Space

An Efficient Spectral Leakage Filtering for IEEE 802.11af in TV White Space Phu Xuan Nguyen 1 , Thinh Hung Pham 2 , Trang Hoan g 2 , and Oh -Soon Shin 3* 1 School of Electronic E ngineering, Soongsil Univers ity, Korea 2 Department of Electrical – Electro nic Engineerin g, Ho Chi Minh City Uni versity of Technology, Vietna m 3 School of Electronic E ngineering a nd Depart ment of ICMC Convergence T echnology, So ongsil Univers ity, Korea * Corresponding author (E -ma il: osshin@ssu.ac.kr) Abstract — Orthogo nal frequency division multiplexing (OFDM) has been widely adopted for modern wireless standards and become a key enabling technology for cognitive radios. However, one of its main drawback s is significant spectral leakage due to the accumulation of multiple sinc -shaped s ubcarrie rs . In thi s paper, we present a novel pulse shap ing scheme for efficient spectra l leakage suppression in OFDM bas ed physical layer of IEEE 802.11af standard . With convent ional pulse s haping filters such a s a raised-cosine filter, vestigial symmetry can be used t o reduce s pectral lea kage very effectively. How ever, these pulse shaping filters require long guard interv al , i.e., cyclic pref ix in an OFDM system, to avoid inter-symbol interference (ISI), resulting in a loss of spectral efficien cy . The pro posed pulse shaping method based o n asymmetric pulse shaping achieves better spectral leakage suppres sion and decreases ISI ca used by filtering a s compared to conventional pulse shaping filters. 1 I. I NTRODUCTI ON Orthogona l frequency division multip lexing (OFDM) is being widely used for many wireless comm unication system s as well as Cognitiv e Radios (CR) due to many advantages [1] . OFDM divides a wideband channe l into na rrow band flat- fading subchannels. Inter-symbol interferenc e (ISI) caused by multipath propagati on can be elim inated by insert ing a cyclic prefix (C P) into each OFDM sym bol. Moreover , OFDM makes efficient use of spectrum by allow ing overlapping of subcarrie rs. However, OFDM has several drawbacks , one of which is s pectra l leakage caused by the accumulation of multiple sync - shaped subcar riers [2]. T his problem can be resolved by the us e of a w ell-desig ned puls e shapin g filt er . To reduce spectral leakage , it is necessary to increase the roll-off fact or of the p ulse shaping filter. Conventiona l pulse shaping filters restrict the roll-off region to b e sm aller than the length of the CP to prevent ISI. This implies that an increase in 1 This work was supported in part b y Institute for Information & Communications Technology Promotion (II TP) grant funded by the Korean government (MSIT) (No. 2017-0-00724, Developme nt of Beyond 5G Mobile Communication T echnolog ies (Ultra-Reliable , Low-Late ncy, and Massive Connectivity ) and Combined Access Technologie s for Cellular-based Industrial Automation Systems), i n part by B asic Science Research Program through the National Resear ch F oundation of Korea (NRF) funded by the Ministry o f Education (No. 2017R1D1A 1B03030436), and in part by the NRF funded by the Ministry of Science and I CT (No. 2017R1A5A 1015596). the roll-off facto r necess itate s an increase in th e CP length , deteri orating the spect ral efficien cy [3], [4], [5]. Recent ly , developm ent of 5G sys tem requires an OFD M sy stem with high spectral efficiency without inc reasin g the CP overhe ad. As ymmetric p ulse shaping filter was introduced to achieve such a goal. We propose the use of an asymmetric pulse shaping filter for IEEE 802.11af standa rd, which is based on OFDM and operates in TV white space (TVWS). Note that TVWS is a portion o f the spectrum not used b y TV broadcast ing in VHF and UHF terrestr ial TV bands [6] . IEEE 802.11af is a global standar d for wirel ess local area network (WLAN) that uses CR technol ogy to operate in the TVWS [7]. The PHY layer config uration of the IEEE 802.11af is defined based on the VHT (very high throughput ) 40 MHz mode of the IEEE 802. 11ac [6] , [8]. One problem to be resolv ed is to obtain the 55dB attenuation requiremen t of its spectrum emission mask (SEM) [7]. In this p aper, we present a novel pulse shaping schem e for efficient spectral leakage suppressi on in OFDM based physical layer of IEEE 802.11af. T he proposed pulse shaping method is based on asym metric pu lse s haping and extensi on of gua rd band. To th e best of our knowledge, there is no publ ished filtering metho d that can suppress spect ral leakage of 802.11af signal with a short guard inte rval. The main contributi ons of this pape r are s umm arized as f ollow s.  The paper presents a novel filtering schem e for 802.11af sy stems that use a short guard interv al , i.e., CP , not to d egrade the spectral efficiency . Using the method, the spectru m leakage of the IEEE 802.11af can be low enough to m eet th e SEM .  The use of an asymmetric pulse shaping can overcome the limited smoothing duration constraine d by a shor t CP .  Extending frequen cy guard interpol ation by increasing the num ber of FFT is presen ted. T his s ignificant ly limits the effect of image spectra, leading to reduced FIR filte r length . The rest of this paper is organized as f ollow s. Section II describes an OFDM system model and introdu ces IEEE 802.11af standard. In Section III, we present the propose d pulse shaping scheme. Numerical results are presented and discussed in Section IV. Finally , conclusi ons are drawn in Section V. II. S YSTEM M ODEL The trans mit sig nal x ( m ) of an OFDM sym bol is defin ed as  󰇛  󰇜      󰇛  󰇜     󰇛   󰇜             (1) where N is the length o f th e invers e f ast Fou rier transfo rm (IFFT), N CP is the length of the CP, and         denotes the overall length of an OFDM symbol.  󰇛  󰇜 denotes the transm it symbol on the   subcarrier , and  󰇛  󰇜 denotes the   sample of an OFDM symbol. Each OFDM symbol can be shaped by a time waveform (pulse shaping wavef orm) . In this case , the OFDM transm it sig nal can be w ritten as  󰇛  󰇜       󰇛  󰇜  󰇛      󰇜     󰇛     󰇜     (2) where the waveform  󰇛     󰇜 determ ines the p ulse shap e. Note that the convention al OFDM signal in (1) corresponds to the case of a rec tangula r wavef or m, i .e .,  󰇛  󰇜                (3) The IEEE 802.11af has defined a television very high- throughput (TV HT) physical-l ayer specifications for the basic channel units (BCUs) of 6, 7, and 8MHz. T he sampling clocks are modified to match to each o f the B CU bandwidths. Table I show s the major param eters of IEEE 802.11af. It inclu des 6 M Hz , 7 M Hz , and 8 M Hz bandw idth channels. 6 M Hz and 8 M Hz bandw idth channels contain 144 s ub -carriers , which comprises of 108 data subcar riers, 6 p ilot subc arriers , and 30 null subcarri ers. A 7 MHz bandwidth channel conta ins 168 subcarrie rs, which comprises of 108 data subcar riers, 6 pilot subcarrie rs, and 54 null su bcarriers . Subcarrie r indexes correspon ding to data subcarriers and pilot subc arriers are as follow s:  Data subcarriers: -58 to -2, 2 to 5 8  Pilot carrier s: -53, -25, -11, +11, +25, +53 The guard interva l is   for 6 M Hz and 7 M Hz bandwidth channels, and     for 8 M Hz bandwidth channel. In IEEE 802.11af, the maximum delay sp read, i.e ., the length of the channel impulse response (CIR) is   [9] . Therefo re, guard interval in Table I can pr event OFDM sy mbols fr om ISI . In this paper, we w ill co nside r the shortest guar d interval (and hence possibly the most problematic case) corresponding to the 8MHz B CU to investig ate the performan ce of the proposed filtering method for 80 2.11af. III. P ROPOSED P UL SE S HAPI NG S CHEME This section proposes a novel method to achieve the SEM requirement o f 802.11af in the shor test guard interval mode . The proposed method is b ased on spect rum shaping, using an TABLE I . T IMING -R ELATED P ARAMETERS OF 802.11 AF FOR TVHT Parameters 6 MHz 7 MHz 8 MHz Number of data s u bcarriers 108 108 108 Number of pilo t subcarriers 6 6 6 Total number of subcarriers 114 168 114 Highest data su bcarrier index 58 58 58 Subcarriers fre quency spacing       IFFT/FFT p eriod   24  18  Guard interval duration  3  2.25  asymm etr ic pulse w ith extended smoothing edge duration, extended guard band, and an FIR filte r . Specific ally, the proposed method works as the follow ing steps: First, a pulse shaping is employed to shape the spectral leakage below the require d pow er. An asymm et ric pulse is a dopted to achieve longer smo othing ed ge d urati on, wh ich leads to larger spectral leakage sup pression w hile minim izing the ISI. This step will be described in detail in Section III-A. Second, the guar d band is extended [2], [10] by incre asing the IFFT size to M tim es the original IFFT size N. The sampling frequency is also increased by M times to maintain the same subcarrier spacing. T his step widens the gap b etween the main spectrum and adjacent image spectra , which allow s a short FIR filter to cancel the image spectra. This ste p is ela borated in Section III-B. A. Pu lse Shaping By smoothing the edges of rectangu lar pulse, the spectral leakage can be reduced [11]. One soluti o n to o btain this is to apply w indowing after appending CP and cyclic suffix (CS) before and after each OFDM sym b ol, respectively , as show n in Fig 1. Fig. 1 . P ulse shaping operation on OFDM symbols. In general , an increase in the roll-off factor β of the window ing function leads to a decrease in the spect r al leakage . However, smoothing edge duration β N T is restricted to be less than the length o f the CP to pre vent channel-induce d ISI [2], [3], [12]. We now investigate three d ifferen t pulse shap es. The first smoothing function in 󰇛  󰇜 , denoted as   , is raised cosine function, recomm ended in the IEEE 802.11 a standar d:   󰇛  󰇜                                                                               (4) The second, denote d as   , is based on the chara cteristics of function w ith vestigial sym metry as derived in [13]. The function of the ves tigial symmetry window is given as   󰇛  󰇜                                                                                                    󰇛    󰇜   (5) As mentioned in the previous section, asymmetric window can provide a trade- off between spectral efficiency and I SI. This window causes lower ISI in com parison to the symmetric window s [3]. The function of an asymmetric window can be written as   󰇛  󰇜        󰇧  󰇛      󰇜  󰇛     󰇜 󰇨                      󰇧  󰇛     󰇜  󰇛     󰇜 󰇨       󰇛    󰇜    (6) Fig . 2 illustr ates the shaped spectra using of different pulse shaping w indows. RC n , VSn , and ASn denote the shaped spectrum using raised cosines functi on, vestig ial sy mm etry function and asymm etr ic puls e, r espectively , with t he smoothin g edge durat ion equal to the length of n original samples. Note that the proposed meth od employs the asymm etr ic pulse to obtain a longer smooth ing edge duration. RC4 and AS4 have almost the same spectral leakag e level that is slightly b etter than VS4 and signific antly better than RC1 . In the pr oposed m ethod, the as ymmetric pulse is empl o yed to extend the smoothing edge duration while keeping the ISI at acceptabl e level. As can be seen in Fig. 2 , AS16 denoted the pulse shaping of the proposed method achieves a m uch better shaped spectrum compared to RC4 . T hat gives the possibility to filter the spectrum leakage to meet the SEM requiremen t of 802.11af. According to the filterin g scheme presented in [14] and the parameters of 8 MHz channel in Table I , the smoothing edge duration β N T should be conservative ly chosen such that it is less th an the durati on of 4-sa mples to avoid ISI . B. Exten ding Frequen cy Guard and FIR Filter In the above s ub section , we do not consi der the existenc e of i ma g e sp e c t r a , a s il l u st r a t ed i n Fi g . 4 fo r t he c a s e o f up - sampling o f 8 times the original frequen cy, wh ich is ca used by the inter polation after the IFFT. Since the guar d band o f the IEEE 8 02.11af is narrow, it is necessary to use inter polation to Fig. 2 . Comparison of shaped spectra with different shaping scheme s. Fig. 3. Comparison of the spectrum after interpolation. increase samplin g frequenc y and expand the baseband bandw idth. Unfortunat ely, interpolati on causes the appeara nce of the imag e spectrum. Becaus e the band gap between the main spectrum and adjacent image spectrum is quite narrow, i t is difficult to cancel the image spectra to avoid inter-channel interfer ence (ICI) . An FIR filter may b e used to elim inate image spectrum [15] [16] [17] , caused by interpolation . However, an FIR filte r can cause ISI because it reduces the effective guard in terval. In th e proposed metho d, the gu ard band is extended by increasing the size of the IFFT to 4 tim es the original IFFT size, i.e.,         . As a result, the proposed method needs the interpolat ion with the factor of 2 instead of 8 to perform up samplin g of 8 times. I n addition, b y doing that the number of imag e spectra decreas es and the gap betw ee n the adjacent spectra and the main spectrum increases. Th is relaxes the r equirem ent of the FIR filter to cancel the image spect ra while m ini mizin g the ISI. Fig. 3 show s the shaped spect rum of th e three meth ods after interpol ation. RC4 , ASp , and Pro denote the shaped spectrum using raised cosines function w ith β N T = 4 * U , asymmetric pulse f unction with β N T = 1 6 * U and th e proposed pulse shaping, respect ively. T he presence of im age spectra limits the spectrum leakage shaping of ASp . T hanks to the extension of the guard band , Pro achi eves ex cellent s pectrum shaping. T abl e II su mmari ze s the mai n de sig n pa ra mete rs o f the proposed method as compar ed to the exiting methods. T he CP is cho sen from sho rt guard inter val mod e t hat eq uals 16* U TABLE II . P ARAMETERS OF T HREE A PPRO ACHES Asymm etric Pulse Shaping State of The Art Method Proposed Method Up Sa mpling ( U ) 8 8 8 NFF T 128 512 512 I nterpola tion 8 2 2 Cycl ic Pre fix 16* U CIR 7* U FI R 9* U 5* U 5* U βN T 16* U 4* U 16* U Fig. 4. The filtered spectrum of the proposed method in comparison with the state of the art methods and SEM. samples after up s ampling . T he channel im pulse response (CIR) equals 1 µ s that is equiv alent to 7 * U sam ples after up sampling . T he param eters are used for simulation to evaluate the perfo rmance of the proposed method th e foll owing Section. IV. N UMERICAL R ESULTS AND D ISCUSSI ON In this Section , we evaluate the filterin g performance of the proposed meth od in terms of spectrum leakage and SER. Two exiting methods are considered for compar ison purpose . One is the conventi onal filterin g based on an asymm etr ic p ulse shaping in [3]. The other is the filtering scheme presented in [14] based on ext ended g uard band. Up sam pling by 8 are investigat ed and the parameters in Table II are used for the three m ethods. Fig. 4 i llustrates the res ult of spectrum le akage filtering . SEM represents the Spectral Emission Mask of IEEE 802.11af. As can be seen in Fig. 4 , ASp has two w id e auxiliary peaks beside the main spectrum and introduces visible distort ion in t h e ma i n s p e c t r u m . D u e t o t h e l i m i t e d l e n g t h , t h e b a n d transiti o n of the FIR filter is not narrow enough to filter out t he image spectra. T he filtere d spectrum of ASp method is far from meeting the SEM requirem ents for 802.11af . The SoA method widens the distance betw ee n the main sp ectru m and adjacent im ag e sp ec tr a d ue to ex te n de d g u ar d b an d . T hi s g iv e s t he possibili ty for using a relative ly short FIR filter to complete ly Fig. 5. The comparison of the SER performance in AWGN. remove the image spectra in the filte red spectrum of SoA . However, the SoA m ethod us es th e sym metric pulse that is constrain ed with a limited roll-of f factor to avoid ISI. Therefore , the side-lo bes of SoA spectrum ar e not low enough to meet the SEM requirem ents. The spect rum leakage of Pro is almost as low as the SEM requirements, as the Pro uses an asymm etr ic pulse to lower ISI that allows longer smoothing edge, i.e., larger roll-off factor for pulse shaping . T his leads to significan t si de-slob suppres sion. In addition, Pro employs extend guard band to cancel the image spectra w ith a s hort length F IR filter . Fig . 5 shows the SEM perform ance in an AWGN channel . Pro is show n to provide almost identical perf orman ce to SoA . This means that an increase in the smoothing edge duration o f the asymmetric pulse in the proposed method cause s neglig ible ISI . Pro achieves better SER than Asp , although both of them use the asymm etric pulse shaping. This is because the lon ger FIR filte r in AS p causes th e larg er ISI. The simu lation r esults show that the proposed method can achieve the spectr al le akage suppression to meet the SEM requirem ent of 802.11af. The ISI caused by performin g filtering in the proposed method is negligible , as verif ied by SER perfo rmance. V. C ONCLUSION This paper presents a novel puls e sha ping scheme for efficient spectral leakage suppressi o n in OFDM based p hysical layer of IEEE 802.11af. T he proposed pulse shaping method is based on an asy mm etric pulse sha ping and extend ed guard band to overcome the constraint o f short guard interv al require d for high spectral eff iciency . The use o f asymm etric pulse shaping reduces the effect of ISI leading to the improve d SER of the proposed method. T he extended gua rd band allows an efficient suppression of image spectra with a short FIR filter. Numerical results are provided to verify the performance of the proposed schem e in terms of the spectral leak age and the sym bol error rate . VI. R EFERENCES [1] T. H. Pham, S. A. Fahmy and I. V. McLoughlin, "An End- to -End Multi-Standar d OFDM Transceiver Architecture Using FPGA Partial Reconfiguratio n," IEEE Access, vol. 5, pp . 21002-21015 , 2017. [2] T. H. Pham, S. A. Fahmy and I. V. Mc Loughlin, "Spectrally Efficient Emission Mas k Shapi ng for OFDM Cognitive Radios," Digital Signal Processing, vol. 50, pp. 150-16 1, 2016. [3] T. T aheri, R. Nilsson a nd J. Beek, "Asymmetric Transmit-Windo wing for Low- La tency and Robust OFDM," in Glob ecom Workshops , 2016. [4] J. Lorca, "Cyclic prefix overhead reduction for lo w- latency wireless co mmunications i n O FDM," in IEEE Vehicular Techn ology Conference (VTC Sprin g) , 2 015. [5] E. Guvenkaya, E. B ala, R. Yang a nd H. Arslan, " T ime- asymmetric and subcarr ier-spec ific pulse shaping in OFDM-based waveforms," IEEE Transactions on Vehicular Techn ology, vol. 64, pp. 5070-5082 , 2015. [6] "IEEE 802 Standard; P art 11; Amendment 5: Television White Spaces (T VWS) Oper ation," IEEE, 2013. [7] Z. Lan, K. Mizutani, G. Villa rdi and H. Harad a, "Design and implementation of a W i-Fi proto type system in TVWS based on IEEE 802.11af," in IEEE Wireless Communication s and Networking Conference (WCNC) , 2013. [8] "IEEE 802 Stan d ard; Part 11; Amendment 6: Wireless Access in Vehicular Environ ments," IEEE, 2010. [9] Z. Lan, K. Mizutani, G. Villa rdi and H. Harad a, "Design and i mplementation of a Wi -Fi pro totype system i n," in IEEE Wireless Communications and Networking Conference (WCNC ) , April, 201 3. [10] T. H. P ham, Techniques for multi-standard cogn itive radios on FP GAs, PhD Thesis, 2015. [11] H. A. Mahmoud and H. Arslan, "Sidelo be suppression in OFDM-based spectrum sharing systems u sing adaptive symbol transition," IEEE Commu nications Letters, vol. 12, pp. 133 -135, 2008. [12] K. Mallikanti, " Analysis of OFDM Syste m b y U sing Pulse S haping Filters for DSP Applications," SSRG Internationa l Journal of Electronics an d Commu nication Engineering, pp. 22-27, 2017. [13] D. Castanheira a nd A. Gameiro, "Novel Windowing Scheme for Cognitive O FDM Systems," IEEE Wireless Communication s Letter, vol. 2, no. 3, p. 251 – 254, 2013. [14] T. H. Pham, I. V. McLoughlin and S. A. Fahmy, "Shaping spectral lea kage fo r I EEE 802 .11p vehicular communications," in IEEE Vehicular Tech nology Conference (VTC) , 20 14. [15] P. Sio han, C. Siclet and N. Laca ille, " Analysis and d esign of OFDM/OQ AM syste ms based on filterbank theor y," IEEE Transactions on Signal Proce ssing, vol. 50, pp . 1170 -1183, 2002. [16] Y. P. Lin, C. C. Li and S. M. Phoong, "A filterbank approach to w i ndow designs for multicarrier systems," IEEE Circuits an d Systems Magazine, vol. 7, pp. 19-30, 2007. [17] C. Muschal lik, "I mproving an OFDM receptio n usin g an adaptive Nyquist windowin g," IEEE Tran sactions on Consumer Electronics, vol. 42, p p. 259 -269, 1996.