THE USE OF CONVOLUTIONAL CODE FOR NARROWBAND INTERFERENCE SUPPRESSION IN OFDM-DVBT SYSTEM Azura Abdullah, Muhammad Sobrun Jaml Jamal, Khazuran Abdullah, Ahmad Fadzl Ismal and An Lza Asnaw Department of Electrcal and Computer Engneerng, Internatonal Islamc Unversty (IIUM), Kuala Lumpur, Malaysa E-Mal: azura.abdullah@gmal.com ABSTRACT The problem of mtgatng narrowband nterference (NBI) due to coexstence between Dgtal Vdeo Broadcastng-Terrestral (DVB-T) and Internatonal Moble Telecommuncaton-Advanced (IMT-A) system s consdered. It s assumed that a spectrum of IMT-A system between 790-862 MHz nterfere the spectrum of the OFDM sgnal n DVB- T band. Two types of convolutonal code (CC) whch s non-systematc convolutonal code (NSCC) and recursve systematc convolutonal code (RSCC) are proposed to mtgate NBI. The performance of the two technques s compared under addtve whte Gaussan nose (AWGN) channel. It s observed that NSCC has a better bt error rate (BER) performance than RSCC. The result showed good performance for low SNR ( 5dB). Keywords: OFDM, convolutonal code, narrowband nterference, DVB-T. INTRODUCTION Orthogonal Frequency Dvson Multplexng (OFDM) s a popular multplexng scheme used for transmsson of hgh data rates n varous communcaton standards such as DVB-T, WLANS, WMANs and WMAX [1, 2]. For certan standards such as WLANs and WMANs, an OFDM system has the ablty to operate n unlcensed frequency bands. As a result, there s a possblty that they have to share the same frequency band wth other communcaton systems such as cordless telephones, garage door openers and baby montors. Ths leads to narrowband nterference (NBI) n the systems [3]. Another example of systems sharng the same frequency band s WMAX and UWB systems. Accordng to [4], UWB system s requred to modfy ts spectrum to avod nterference wth WMAX. In WRC-07 conference, ITU-R has allocated the 790-862MHz frequency for IMT-A system. Ths also suggests that the DVB-T system whch operates between 470-862MHz have to share ts upper frequency band wth the IMT-A system [5]. There are several technques proposed to mtgate NBI such as usng orthogonal codes, frequency doman cancellaton, recever wndowng and excson flterng [3, 4]. Although orthogonal codes s found to gve better performance compared to error control code (ECC), ths method does not comply wth the current OFDM standard such as DVB-T and IEEE [3]. On the other hand, frequency doman cancellaton technque s not sutable to be mplemented n broadcastng because the channel and nterference nformaton from the recever needed to be fed back to the transmtter perodcally for update. A lmtaton of recever wndowng method s that t s sutable to be used together wth frequency doman cancellaton to reduce the effect of snc shape sde lobes from spreadng to adjacent channel whle excson flterng method provdes less beneft wth quadrature ampltude modulaton (QAM) [4]. ECC s a sutable canddate to mtgate NBI as t s able to protect the data usng a specfc code. The data whch s corrupted durng transmsson n the nosy channel wll be recovered by the specfc decodng method. Theoretcally, ECC has the capablty to lower the bt error rate of an uncoded system by a certan codng rate [6]. In bref, there are three types of ECC known as block, convolutonal and modern codes. In ths work, convolutonal code (CC) s proposed as t s sutable to be used n broadcastng, deep space communcaton, dgtal speech and also for Gaussan channel condton [7, 8]. Two types of convolutonal code proposed to mtgate NBI n DVB-T system are non-systematc convolutonal code (NSCC) and recursve systematc convolutonal code (RSCC). Secton 2 descrbes the OFDM system and NBI model used. Secton 3 explans about the proposed ECC technques. Secton 4 provdes the smulaton results and dscusson whle secton 5 concludes ths paper. SYSTEM MODEL The OFDM smulaton model of a DVB-T system under narrowband effect s as shown n Fgure-1. It s referred from a MATLAB smulaton by [9]. The smulaton model s modfed by addng ECC as a narrowband mtgaton technque and usng dfferent carrer frequency. ECC acts as encoder n the transmtter and decode the sgnal back for recovery n the recever. At baseband, ECC s appled at the stream of bnary data k= {k 1 k 2 k 3.. k n }. Then, the coded bnary data c = {c 1 c 2 c 3 c 4 c n }, s converted nto symbols to be modulated wth M number of Quadrature Ampltude Modulaton (QAM). Each seral modulated symbols S = {S 1 S 2 S 3 S 4 S n }, are mapped nto N number of parallel subcarrers. The modulated symbols X(k), appeared as a complex sgnal n frequency doman: X ( k) = R( k) + ji( k) (1) 1183
The modulated symbols are passed to nverse fast Fourer transform (IFFT) processng block to create a tme doman OFDM sgnal for transmsson. 2N-IFFT processng s used to center the subcarrers and processed the dscrete sgnal x(n), N 1 j2 nk N x( k) = 1 N X ( k) e π k = 0 where n = 0,1,, N-1, k = 0,1,2,3 N-1; N beng the number of subcarrers. An OFDM symbol of N subcarrers s to be transmtted n an OFDM symbol perod duraton. The next processng block s to sample the OFDM dscrete sgnal x(n), wthn the OFDM symbol perod duraton. It wll undergo flteraton process n dgtal-to-analog (DAC) converter to obtan contnuous tme doman sgnal x(t). Fnally, the sgnal x(t), s modulated wth ts RF transmt sgnal carrer and ready for transmsson. The recever system s the reverse process of the transmsson system. After demodulaton, a decoder recovered the data based on ECC scheme appled. After the data s recovered, t s compared wth the orgnal data for bt error rate (BER) calculaton. From Fgure-1, the receved sgnal and the effect of channel can be wrtten as follows: r( t) = x( t) + n( t) + ( t) (3) where r(t) s the receved sgnal consst of transmtted sgnal x(t), Gaussan nose (AWGN) n(t) and narrowband nterference (t). Fgure-2 shows the theoretcal model of OFDM- DVBT band adopted from [2] whch s used to represent general scenaro n ths work. For all the channels that are used for transmsson, there are 49 channels n the DVB-T (2) frequency band. From equaton (3), (t) has a frequency range between 790-862 MHz nterfered wth the upper channel n DVB-T band. In ths work, the 48 th channel n the DVB-T band whch has carrer frequency of 850MHz s chosen as smulaton parameter wth the unwanted NBI sgnal of frequency 851MHz. The narrowband nterference (NBI) sgnal s modeled as snusodal sgnal ( t ), ( ) ( ) t = Icos 2π f + θ 0 < θ < 2π (4) where I s the ampltude of the NBI sgnal and s the phase angle. The value of consdered n ths work s θ = π Substtutng (4) nto (1), the receved sgnal r( t ), s derved as: ( ) cos( 2π ) ( ) cos( 2π θ) r t = A f t + n t + I ft+ (5) c where A s the ampltude and f ( c ) s the carrer frequency of the OFDM sgnal. The performance of an OFDM system s degraded when a strong NBI sgnal f wth carrer frequency close to the OFDM sgnal s carrer frequency f ( c) overlapped, f = f + f (6) c and that the ampltude of the NBI sgnal s greater than the ampltude of the OFDM sgnal ( I > A). Further detals can be found n smulaton part-4. Fgure-1. OFDM smulaton model wth DVB-T parameters. 1184
THE PROPOSED TECHNIQUE Fgure-2. Theoretcal model of OFDM-DVBT band adopted from [2]. Non-systematc Convoluton Code (NSCC) Fgure-3 shows the block dagram of 1/2 rate NSCC encoder whch consst of m number of memory regsters. It s used to store prevous bnary nput data. If a bnary data k, enters an encoder, t produces n coded bts at the output wth code rate R=k/n. The code representaton s wrtten as (n,k,m). The desgn of NSCC can be found n lteratures such as [7, 8] and [10] wth dfferent generator polynomals. In ths work, the generator polynomals used are dfferent compared to the ones used n example [7, 8] and [10] because based on smulaton result, t gave BER performance curve closest to the OFDM system wthout NBI effect. The generator polynomals used for 1/2 rate NSCC are g 1 = [1 1 1] and g 2 = [0 1 1]. Fgure-3. 1/2 rate convolutonal encoder (2, 1, 3). An nput bt 1 whch entered the encoder wll be modulo-2 added wth stored values n memory regster to generate the coded bts u 1 u 2. The generator polynomals determne whch stored values n memory regster needed to be modulo-2 added wth the nput bt. Assumng the ntal state of memory regster s 000, the output s shown n Table-1 below. Start state Table-1. Example of truth table for 1/2 rate convolutonal encoder (2, 1, 3). Input End state U1 U2 Output 000 0 000 (0000) = 0 (000)=0 00 000 1 100 (1000) = 1 (100)=0 10 100 0 010 (0100) = 1 (000)=0 10 100 1 110 (1100) = 0 (100)=1 01 The nput bt s then moved nto shft regster m 1 wth all the bts n the memory regster shfted. The oldest stored bt n m 3 s dsappeared. The next nput bt entered wll be modulo-2 added wth stored values n memory regster whch s 100 and the process wll be repeated. The system s then extended to 1/3 rate by addton of another generator polynomal, g 3 = [1 0 1] as shown n Fgure-4. In the case of 1/3 rate encoder, one bt entered the encoder produced three output bts. Fgure-4. 1/3 rate convolutonal encoder (3, 1, 3). 1185
NSCC gave better performance when Vterb decoder s used [8]. The possble path that the encoder has undergone s represented n Trells dagram as shown n Fgure-5. All the memory regster s possble state s wrtten at the frst column. The nput bt (n bracket) s wrtten next to the matched output bts referred to encoder s truth Table. Usually, the process wll start at state 000. The branch metrc s calculated by comparng the number of bt agreement wth the coded bts. The process s repeated for all the coded bts. The path whch has the hghest number of branch metrc s chosen as survvor path and the decoded bts are determned. Coded bts 00 01 Branch metrc State 000 00(0) 00 (0) 3 001 010 10(0) 1 011 100 10 (1) 10 (1) 2 101 110 01(1) 3 111 Fgure-5. Example of Trells dagram for 1/2 rate convolutonal code. Recursve systematc Convoluton Code (RSCC) RSCC, also known as turbo code s developed from NSCC. Compared to NSCC, RSCC s formed by concatenatng n parallel two RSCCs separated by an nterleaver. It s a systematc codng because one of the message bt tself s called systematc bt and the other two are the party bts generated by the two RSCC encoders. Snce the am of ths research s to determne whch type of convolutonal code performs better n mtgatng NBI, the desgn of RSCC used n ths paper s adopted from [11]. Its codng rate R, s 1/3 wth generator polynomal g 1 = [101], g 2 =. [111], and g 3 = [101]. Iteratve decodng s used to decode the message. SIMULATION RESULTS The performance of convolutonal codes n OFDM system under NBI effect s smulated usng MATLAB based on the DVB-T parameters for 2k mode [9]. Fgure-6 shows the llustraton of the OFDM sgnal n the presence of Gaussan effect and unwanted NBI sgnal. The BER performance curve of the OFDM system wth and wthout the presence of NBI s presented n Fgure-7. The ntal OFDM curve (wthout nterference) has average sgnal power value of -10dB. The value obtaned s calculated based on smulaton accordng to [1]. The NBI sgnal s added to the ntal OFDM system and smulated n two condtons.e. wth snusodal ampltude I=10V and I=20V. Referrng to [6], the NBI sgnal power for the case of equaton (4) s, 2 P I = (7) 2 where I s the ampltude of the narrowband snusodal sgnal. The NBI sgnal power for snusodal ampltude I=10V and I=20V are approxmately 17dB and 23dB respectvely. For NBI sgnal power less than 17dB, t has a small effect on the OFDM system. In Fgure-7, at SNR=5dB, the bt error rate s 0.06631 for system under NBI sgnal power of 17dB and 0.1332 for NBI sgnal power of 23dB compared to 0.03815 for ntal system. The dfference of about 0.02816 between system wth 17dB NBI power and ntal system s due to the fact that the system performance s affected by the NBI sgnal. When the NBI s sgnal ampltude s ncreased to I=20, the dfference wth ntal system s 0.09505 whch mples further degradaton of the system performance due to the ncreased n NBI sgnal power. Fgure-8 shows the performance of convolutonal coded (CC) OFDM wth the presence of 17dB NBI sgnal power. For the case of SNR = 5dB, the bt error rate s about 0.04577 for 1/2 rate NSCC whch s close to the ntal system. As the SNR ncreased, the performance of 1/2 rate NSCC dd not follow the curve of ntal OFDM system. RSCC obtaned BER of 0.09507 at SNR=5dB whch showed worst performance compared to the OFDM system wth NBI. On the other hand, 1/3 rate NSCC gave faulty result because t acheved BER of 0.01819 at SNR=5dB whch s lower than the BER of ntal system. The convolutonal coded OFDM under 23 db NBI sgnal power effect s shown n Fgure-9. Based on the fgure, 1/3 rate NSCC outperformed 1/2 rate NSCC and RSCC. At SNR=5dB, an error of 0.04401 s obtaned by 1/3 rate NSCC whch s smlar to the ntal system. As the SNR ncreased, the performance of 1/3 rate NSCC also dd not follow the curve of the ntal system. 1/2 rate NSCC showed less performance as t followed the curve of OFDM system under NBI effect whereas RSCC showed the worst performance wth error rate 0.2019 at SNR=5dB. Based on observatons on Fgure-8 and Fgure-9, at low SNRs (5dB), the performance curve of NSCC followed the curves of ntal system (wthout nterference) compared to hgh SNR (>5dB). Ths mples that the performance of NSCC wth Vterb decoder s dfferent at low and hgh SNR. Accordng to [7], for a convolutonal code, the error correcton and detecton capablty t, s ( d free 1) t = (8) 2 Where d s free dstance whch s the smallest free Hammng dstance between all possble code sequences of the code? At hgh SNR, the performance s lmted by the capablty of Vterb decoder to correct more than t number of errors n n bts. Power, bandwdth constrant and nature of nose n the channel can also affect the performance of the codng scheme. Based on the smulaton results, the proposed NSCC has good performance for low SNR (5dB) to mtgate NBI compared to RSCC. The performance of ths code also 1186
showed consderable result at low SNR when compared wth tme wndowng method for NBI mtgaton under DVB-T system [9]. Power/frequency (db/hz) Power/frequency (db/hz) -80-100 -120-140 Welch Power Spectral Densty Estmate -160 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Frequency (GHz) Welch Power Spectral Densty Estmate -40-60 -80-100 -120 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Frequency (GHz) Fgure-6. An OFDM sgnal spectrum appears between 0.4MHz-1.25MHz along the x-axs for 850MHz carrer frequency, showng OFDM transmtted sgnal spectrum (top) and OFDM receved sgnal spectrum wth Gaussan nose and the presence of 851 MHz NBI sgnal (bottom). 10 0 Y: 0.1332 10-1 Y: 0.06631 Bt Error Rate 10-2 Y: 0.03815 AWGN 16QAM-OFDM AWGN 16QAM-OFDM + (t), I=10 10-3 AWGN 16QAM-OFDM + (t), I=20 0 1 2 3 4 5 6 7 8 SNR n db Fgure-7. BER performance of OFDM system wth NBI effects (blue) and ntal system wthout nterference (red). The smulaton also shows comparson of the system performance when the unwanted snusodal ampltude s vared (I=10V and I=20V), havng dfferent sgnal power. 1187
10 0 10-1 Y: 0.09507 Bt Error Rate 10-2 Y: 0.04577 Y: 0.01819 10-3 AWGN 16QAM-OFDM AWGN 16QAM-OFDM + (t), I=10 AWGN 16QAM-OFDM + (t) + 1/2 rate NSCC AWGN 16QAM-OFDM + (t) + 1/3 rate NSCC AWGN 16QAM-OFDM + (t)+ RSCC 0 1 2 3 4 5 6 7 8 SNR n db Fgure-8. Performance comparson of convolutonal coded OFDM (blue) wth the presence of 17dB NBI sgnal power wth uncoded system (red). 10 0 Y: 0.2019 Bt Error Rate 10-1 10-2 Y: 0.04401 Y: 0.1467 10-3 AWGN 16QAM-OFDM AWGN 16QAM-OFDM + (t), I=20 AWGN 16QAM-OFDM + (t) + 1/2 NSCC AWGN 16QAM-OFDM + (t) + 1/3 NSCC AWGN 16QAM-OFDM + (t) + RSCC 0 1 2 3 4 5 6 7 8 SNR n db Fgure-9. Performance comparson of convolutonal coded OFDM (blue) wth the presence of 23dB NBI sgnal power wth uncoded system (red). 1188
CONCLUSIONS In ths paper, a narrowband mtgaton technque s proposed usng convolutonal code. NSCC and RSCC are presented as two dfferent ECC technques sutable for NBI mtgaton for DVB-T transmsson under Gaussan channel. The nterference s assumed comng from IMT-A servces affected the upper channel of the DVB-T band. The smulaton showed that 1/2 rate NSCC can mtgate the 17dB NBI sgnal power whle 1/3 rate NSCC suted for 23dB NBI sgnal power at low SNR ( 5dB). The performance result for RSCC showed that t s not effectve n mtgatng the NBI for ths work. REFERENCES [1] Y. S. Cho, J. Km, W. Y. Yang, and Chung-Gu Kang, MIMO-OFDM Wreless Communcatons wth MATLAB. Sngapore: John Wley and Sons. [2] E. B. Unon, Ets en 301 701, 2000. [3] A. J. Coulson, Bt error rate performance of OFDM n narrowband nterference wth excson flterng, IEEE Trans. Wrel. Commun., vol. 5, no. 9, pp. 2484-2492, 2006. [4] A. Batra and J. R. Zedler, Narrowband Interference Mtgaton n OFDM Systems, Ml. Commun. Conf. MILCOM 2008, pp. 1-7, 2008. [5] Z. A. Shamsan, T. A. Rahman and A. M. Al-hetar, Pont-Pont Fxed Wreless And Broadcastng Servces Coexstence wth IMT-Advanced System, Prog. Electromagn. Res. vol. 122, no. November 2011, pp. 537-555, 2012. [6] B. P. Lath, Modern Dgtal and Analog Communcaton Systems, New York: Oxford Unversty Press, 1998. [7] Y. Jang, A practcal gude to error control codng usng MATLAB, Boston, London: Artech House, 2010. [8] P. Sweeney, Error Control Codng from Theory to Practce, London, England: John Wley and Sons Ltd., 2004. [9] K. Abdullah, Interference Mtgaton Technques for Wreless OFDM, PhD dssertaton, RMIT, Melbourne, 2009. [10] N. Arshad and A. Bast, Implementaton and Analyss of Convolutonal Codes, Int. Journal of Multdscp. Sc. Eng., vol. 3, no. 8, pp. 9-12, 2012. [11] N. Arshad and M. A. Jamal, Implementaton and Analyss of Turbo Codes Usng MATLAB, Journal of Expert Systems (JES), vol. 2, No. 1, pp. 115-118, 2013. 1189