<B>

Available online at www.sciencedirect.com

ScienceDirect

CrossMark

ELSEVIER ICT Express 1 (2015) 138-142

www.elsevier.com/locate/icte

Tx scenario analysis of FBMC based LDM system

Soonki Jo, Jong-Soo Seo*

Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea Received 28 August 2015; accepted 9 November 2015 Available online 23 November 2015

Abstract

Filter bank multiple carrier (FBMC) technology is one of the alternative solutions for multicarrier modulation. FBMC does not need cyclic prefix (CP) and guard band utilized for orthogonal frequency division multiplexing (OFDM) and CP and guard band cause loss of spectral efficiency. FBMC features offset QAM (OQAM) and band-limited filtering on each subcarrier, which eliminate the need of CP and guard band. FBMC filtering could maintain the orthogonality in a real signal domain by using well localized filter. Meanwhile layer division multiplexing (LDM) is also introduced to increase the spectral efficiency. In LDM, low density parity check (LDPC) coded multi signals are transmitted simultaneously with different power levels in same frequency band and these signals form signal layers. Combination of FBMC and LDM techniques can maximize the spectral efficiency. In this paper, LDM system which adopts FBMC is proposed. The LDM system has both OFDM and FBMC modulated layers. To apply FBMC to LDM system, log-likelihood ratio (LLR) calculation scheme for FBMC is needed for LDPC decoding. Three scenarios for LDM system are considered and BER performance of each scenario is analyzed to find proper scenario. © 2015 The Korean Institute of Communications Information Sciences. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: FBMC; LDM; LDPC; LLR

1. Introduction

Filter bank multicarrier (FBMC) is proposed [1] for alternative solution of multicarrier modulation such as orthogonal frequency division multiplexing (OFDM). FBMC does not need cyclic prefix (CP) and guard interval due to filtering on each subcarrier [2]. In FBMC, Offset-QAM (OQAM) is used because orthogonality of filter is satisfied in real domain. In OQAM, real and imaginary parts of QAM symbols are transmitted separately. Therefore FBMC can achieve high spectral efficiency [3]. As using well localized waveforms in time and frequency domain, the single channel coefficient per subcarrier equalizer structure can be considered.

Another candidate technology for achieving high data rate, termed as cloud transmission network (Cloud-Txn) with layer division multiplexing (LDM) was proposed [4]. LDM is a similar technique with non-orthogonal multiple access (NOMA)

* Corresponding author.

E-mail addresses: joangel21@yonsei.ac.kr (S. Jo), jsseo@yonsei.ac.kr (J.-S. Seo).

Peer review under responsibility of The Korean Institute of Communications Information Sciences.

* This paper is part of a special issue entitled "Next Generation (5G/6G) Mobile Communications" guest-edited by Prof. Jungwoo Lee, Dr. Sumei Sun, Prof. Huaping Liu, Prof. Seong-Lyun Kim and Prof. Wan Choi.

but, the LDM focuses more on multiplexing than multiple access. Multiple signals are transmitted simultaneously on the same frequency with different power. Each signal delivers different service. For example in a 2-layer LDM system, the upper layer delivers mobile services and the lower layer is used to serve (U)HDTV. This system operates with strong forward error correction (FEC) coding, such as low density parity check (LDPC) code.

FBMC and LDM are techniques which can increase the spectral efficiency. FBMC does not require CP and guard band, and LDM uses frequency band with superposing signals. The combination of these two techniques can achieve maximum spectral efficiency, however, FBMC application to LDM system has not been researched, so far. And also, proper transmission scenario for this application has not been introduced.

In this paper, LDM system which adopts FBMC is proposed. Proposed LDM system is two layered system which has both FBMC and OFDM layers, and three transmission scenarios are considered. (1) FBMC over OFDM, (2) OFDM over FBMC and (3) FBMC over FBMC. In the scenarios 1 and 2, OFDM layer is considered as conventional service signal and FBMC signal layer is added. In the scenario 3, both layers consist of FBMC signals. To insert FBMC layer in LDM system, FBMC signal should be decoded with LDPC code. So we also introduce log-likelihood ratio (LLR) calculation method for FBMC [5].

http://dx.doi.org/10.1016/j.icte.2015.11.001

2405-9595/© 2015 The Korean Institute of Communications Information Sciences. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

The organization of the paper is as follows. Section 2 describes FBMC modulation and system model. In Section 3, LLR calculation for FBMC signal is presented. In Section 4, three LDM transmission scenarios are proposed and bit-error rate (BER) performances of scenarios are analyzed. Finally in Section 5, we draw conclusion.

Notation. (•)* denotes the complex conjugate operation.

2. FBMC system

We can represent the FBMC signal at the transmitter as [1]

ej M (m-#) ej

s[m] = ^2dktnp[m = nM/2] k=0 nsZ

where dknp[m] and M are real-valued transmitted OQAM symbol, the prototype filter, and a number of subcarriers, respectively. D is the filter delay term and $k,n is an additional phase term. k and n are indexes for subcarrier, and time instant, respectively. We can rewrite (1) as

stm] = ^ ^2dk,np[m]k,n

k=0 nsZ

where p[m]kn = p[m - nM/2]ej~Mr(m-D)ej$k,n, which is time and frequency shifted version of p[m]. In the receiver, the signal in subcarrier k at time n is determined with the inner product of s[m] and p[m]kn

rk',n' = (s, Pk',n'> = s[m]p*[m]k',n'

+œ M-1

= IZ IZdk,nP[m]k,nP*[m]k',n'.

m=-œ k=0 nsZ

The prototype filter p[m] satisfies the real orthogonality condition given by [3]

Re I ^2 P[m]k,nP*[m]k',n' | = $k,k'$n,n'. [m=-œ J

From (4), (3) can be rewritten as

Ik n ■ intrinsic interference

rk,n = dk,n + m dk',n' ^2 p[m]k,nP*[m]k'y . (5)

k'=kn'=n m=-œ

According to the real orthogonality conditioning (4), the interference Ik,n is imaginary term.

Let us simplify P[m\nP*[m]k'n as ci,m [6]. The

coefficient

ci,m represents the system impulse response, and K is the overlapping factor of the prototype filter. We use K = 4 in this paper, and if dknej$k,n is considered as dkn, the main coefficients are given in Table 1 [7]. Then, jukn is expressed as

1 2K-1

Ik,n = ^2 X Cl,mdk+l,n+m; I, m = 0. (6)

l=-1 m=-(2K-1)

Table 1

Transmultiplexer impulse response cim.

■ n - 2 n-1 n n + 1 n + 2 ■■■

k - 1 •• ■ -0.125 - j 0.206 0.239 j0.206 -0.125 ■ ■ ■

k ■■ ■ 0 0.564 1 0.564 0 ■■■

k + 1 •• ■ -0.125 j 0.206 0.239 - j0.206 -0.125 ■ ■ ■

If the channel is considered as constant during the summation period (-1 < l < 1, -(2K - 1) < m < 2K - 1), and received signal can be written as [8]

rk,n = hk,n (dk,n + Ik,n ) + nk,n.

3. LLR calculation for FBMC

To utilize LDPC code, LLR should be calculated. The efficient LLR calculation scheme for FBMC system has been introduced [5]. The brief description of LLR estimation algorithm of [5] is followed. The advanced work for LLR calculation is necessary. By multiplying h*k n to (7) and operating real value extraction, then received signal can be rewritten as

Re {h*knn,n} = Re {|hk,n |2(dk,n + Ik,n ) + h^M} = |hk,n|2dk,n + Re {h'knnk,n). The LLR is depicted as

£ P(rId)

ttmu^ i pr(bi = 0|r ) x ex0 LLR(bi ) = log —-— = log

Pr(bi = 1|r) E P(rId)

where bl is lth bit, and x0 and x0 denote subsets in which the ith bit of the symbol is 0 and 1, respectively. P(r \d) is the conditional probability density function (PDF). To calculate LLR, variance of noise term should be estimated. In (8), Re{h| nnk,n} is noise, and the noise variance is as follows

Var {Re {h^nu}} = = |hk,n I2a2, where a2 is the variance of noise term nk,n

3.1. Real and imaginary FBMC symbols pass through same channel

If it is assumed that, real d(r)k,n and imaginary jd^)k,n FBMC symbols experience same channel (hk,n = hk'= h), the PDF of the combined QAM symbols is represented as

P(r|d, |h|2) 1

(?(r)k,n - |h|2d(r)k,n)

+ (r(i )k' ,n' - I h 12 d(i )k' ,n')

-1 0 1 Subcarrier index

10 0 -10 -20 -30 -40 -50 -60 -70

-1 0 1 Subcarrier index

Fig. 1. (a) Waveforms of Sinc filter and PHYDYAS filter, (b) Power spectrum of the filters. (Blue: OFDM, Red: FBMC). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.2. Real and imaginary FBMC symbols pass through different channels

If it is assumed that, real d(r)k,n and imaginary jd(i)k,n FBMC symbols experience different channels (hk,n = h1 = ht,n' = h2), the PDF of the combined QAM symbols is represented as

P(r\d, |h 112, |h2|2)

2ncr1cr2

r(r)k,n - \h 1|2d(r)k,n , 2à?

%)k',n' - |h2|2d(i)k'

4. FBMC based LDM system

4.1. Active subcarriers in OFDM and FBMC

FBMC uses the PHDYAS filter which has low sidelobe and maximizes spectral efficiency with no guard band. In this part, we compare sidelobes of OFDM and FBMC. DVB-T2 defines active subcarrier numbers for each FFT mode. For example, in 1 K mode, 852 active carriers are used. In general, for any transmission mode of DVB-T2, only about 85% of the total subcarriers are used for transmission. The remaining 15% of subcarriers are used as guard band. Without this guard band, high sidelobe power of OFDM signal interferes with the neighboring services. In Fig. 1(a), the sinc filter of OFDM and the PHYDYAS filter of FBMC are represented. Fig. 1(b) shows filter shapes in dB scale. It is obvious that sidelobes of FBMC are markedly lower than that of OFDM. The power of first sidelobe of sinc filter is -17 dB, however the first sidelobe of FBMC is almost -40 dB. This distinction between two modulations becomes remarkable in the multicarrier environment.

Fig. 2 shows power spectral densities (PSD) of OFDM and FBMC in 1 K mode. As DVB-T2 stipulates, 86 null subcarriers

Fig. 2. PSDs of OFDM and FBMC (1 K mode and 86 Guard bands on each side).

are allocated on both sides. An interesting observation that can be made is that the OFDM sidelobe power merely reduces to approximately about -55 dB on its either end whereas it only requires a single null subcarrier on either end of FBMC to reduce sidelobe power to approximately -60 dB.

In this paper, we consider 1 K mode of transmission with number of null subcarriers for OFDM and FBMC as 172 and zero, respectively.

4.2. LDM scenarios

In the FBMC and OFDM combined LDM system, we consider three transmission scenarios.

(1) Scenario 1. FBMC over OFDM.

Fig. 3(a) represents LDM scenario 1. In scenario 1, FBMC signal is transmitted in the upper layer and OFDM signal is transmitted in the lower layer. OFDM signal in existing service is overlaid with FBMC signal.

(2) Scenario 2. OFDM over FBMC.

Fig. 3(b) represents LDM scenario 2. In scenario 2, OFDM signal is transmitted in the upper layer and FBMC signal is

Fig. 3. (a) LDM scenario 1. (b) LDM scenario 2. (c) LDM scenario 3.

transmitted in the lower layer. Underlying FBMC signal is transmitted under OFDM signal.

(3) Scenario 3. FBMC over FBMC.

Fig. 3(c) represents LDM scenario 3. In scenario 3, both upper and lower layers have FBMC signals and it is assumed that conventional service also uses FBMC modulation.

5. Performance analysis

In this section, performances of upper layers in the additive white Gaussian noise (AWGN) channel, are analyzed. Because if upper layer signal is demodulated perfectly, lower layer signal can be easily detected with canceling upper layer signal. Figs. 4-6 represent BER performance of each scenario according to power gap between upper layer and lower layer. F/O, O/F and F/F are 'FBMC over OFDM' (Scenario 1), 'OFDM over FBMC' (Scenario 2), and 'FBMC over FBMC' (Scenario 3), respectively. QPSK is used for simulation and considered FEC code is LDPC code with code rate of 1/3, 1/2, and 3/5.

The performances of scenarios 1 and 2 are very similar, and scenario 1 looks a little bit better. For the scenario 3, more power is necessary about SNR 1-2 dB to reach same performance of scenario 1. In the scenario 1, lower OFDM layer signal interferes with upper FBMC layer, but because of null subcarrier, some FBMC signals are distorted less. In the scenario 2, all upper OFDM layers are affected by lower FBMC layer. In the scenario 3, all upper FBMC signals are interfered by lower layer that cause the highest performance degradation.

Meanwhile, we can discover the interesting result when the code rate is 3/5. In the 1 dB power gap case, BER floors appeared in scenario 1 and scenario 2. Scenario 3 represents good performance, but also has BER floor at 10-61 at SNR 18.7 dB. In the 3 dB power gap case, the BER performances of scenarios 1 and 2 become similar to that of scenario 3. When the lower layer has 5 dB low power than upper layer, performances of scenarios 1 and 2 surpass that of scenario 3.

6. Conclusion

In this paper, we considered LDM system based FBMC. Proposed LDM system consists of two layers, and three transmission scenarios are considered, which are (1) FBMC over OFDM, (2) OFDM over FBMC, and (3) FBMC over FBMC. For organizing signal layers, PSDs of OFDM and FBMC signal are analyzed. The sidelobe power of OFDM signal is much higher than that of FBMC signal. Guard band is allocated only for OFDM. Because of the distinct application of guard band, the performance of LDM in each scenario has different tendency. Scenario 1 outperforms other scenarios in

Fig. 4. BER performance of 1st layer with 1 dB power gap between two layers.

Fig. 5. BER performance of 1st layer with 3 dB power gap between two layers.

Fig. 6. BER performance of 1st layer with 5 dB power gap between two layers.

almost all the cases. Scenario 2 has similar performance to that of scenario 1. But in the case of 1 dB gap between layers with code rate 3/5, scenario 3 has best performance.

From this analysis, we can observe that LDM system with the combination of FBMC and OFDM is appropriate. Combined LDM system has better performance, especially FBMC over OFDM case, rather than LDM system composed with only FBMC signals.

Acknowledgment

This work was supported by the ICT R&D program of MSIP/IITP (1391202006, A Study on Next Generation Interactive Terrestrial Broadcasting System).

References

[1] B. Le Foch, M. Alard, C. Berrou, Coded orthogonal frequency division multiplex, Proc. IEEE 83 (6) (1995).

[2] M.G. Bellanger, Specification and design of a prototype filter for filter bank based multicarrier transmission, in: Proc. IEEE ICASSP, Salt Lake City, USA, May 2001, pp. 2417-2420.

[3] P. Siohan, C. Siclet, N. Lacaille, Analysis and design of OFDM/OQAM systems based on filter bank theory, IEEE Trans. Signal Process. 50 (5) (2002) 1170-1183.

[4] Y. Wu, B. Rong, K. Salehian, G. Gagnon, Cloud transmission: A new spectrum-reuse friendly digital terrestrial broadcasting transmission system, IEEE Trans. Broadcast. 58 (3) (2012) 329-337.

[5] S. Jo, J. Seo, Efficient LLR calculation for FBMC, IEEE Commun. Lett. (2015) http://dx.doi.org/10.1109/LC0MM.2015.2467376.

[6] R. Zakaria, D. Le Ruyet, A novel FBMC scheme for spatial multiplexing with maximum likelihood detection, in: Proc. IEEE ISWCS, Sep. 2010, pp. 461-465.

[7] M. Bellanger, and PHYDYAS team, FBMC physical layer: a primer, website: www.ict-phydyas.org.

[8] C. Lele, J. Javaudin, R. Legouable, A. Skrzypczak, P. Sioha, Channel estimation methods for Preamble-based OFDM/OQAM Modulation, in: Proc. European Wireless, Apr. 2007.