Correlated Scrambling Scheme For Time-frequency Diversity In Ofdm Single-frequency-network Systems

Correlated Scrambling Scheme for Time-Frequency Diversity in OFDM Single-Frequency-Network Systems Hsien-Wen Chang, Chorng-Ren Sheu, Ming-Chien Tseng, and Ching-Yung Chen Information & Communications Research Laboratories, Industrial Technology Research Institute 195, Sec. 4, Chung Hsing Rd., Chutung, Hsinchu, Taiwan, R.O.C. {seanchang, crsheu, mctseng}@itri.org.tw Department of Computer and Communication Engineering, National Kaohsiung First University of Science and Technology No 2 Jhuoyue Rd., Nanzih District, Kaohsiung City, Taiwan, R.O.C. [email protected] Abstract － For orthogonal frequency division multiplexingbased single frequency network systems, such as the terrestrial digital broadcasting systems, the performance of a static/quasistatic receiver at cell edge can be severely degraded by the slow and flat fading effect due to a destructive combination of signals from adjacent transmitters. Recently, the cyclic delay diversity (CDD) and group-wise scrambling diversity (GSD) schemes have been presented to address this issue. However, both the CDD and GSD schemes provide no time diversity against the slow fading nature. Moreover, discontinuities of composite channel responses inherent in the GSD scheme make channel estimation even more difficult. In this paper, a new correlated scrambling diversity (CSD) scheme is proposed that introduces both time and frequency diversities via the time-frequency scrambling independently among transmitters. Continuity of composite channel responses resulting from utilization of correlated scrambling patterns further makes channel estimation easy for receivers. Index Terms—OFDM, SFN, time-frequency diversity

I. INTRODUCTION

A

frequency network (SFN) is a broadcast network where several transmitters send the same signal over the same frequency channel simultaneously [1]. Compared to traditional multi-frequency network (MFN), SFN has some advantages including: 1. Efficient utilization of radio spectrum; 2. Large coverage area due to multiple small-power base stations (BS); 3. User equipments (UE) do not need to switch frequency while wandering in SFN coverage. However, SFN has severe multipath propagation since each UE receives several echoes of the same signal, and the constructive or destructive interference among those echoes typically result in frequency selective fading channels. Coded orthogonal frequency division multiplexing (COFDM), a multicarrier modulation technique in conjunction with channel coding, is effective in processing frequency selective fading channels [2]. The frequency selective fading effect is treated as the resource of diversity for COFDM systems. Signals severely corrupted at subcarriers experiencing deep fading can be recovered by signals at other subcarriers with good channel condition as long as the amount and pattern of errors are within capability of the utilized channel coding/interleaving scheme. Therefore, COFDMbased SFN architecture has been extensively adopted by SINGLE

terrestrial broadcasting systems such as DVB-T/H, DAB and T-DMB, etc. for broadband wireless transmission in a complicated multipath environment [3]-[6]. At the cell edge between transmitters in an SFN, it may happen that a receiver receives the same signal from the two transmitters almost simultaneously. The tiny delay spread results in flat fading channel response with a wide coherent bandwidth [7]. If signals from the two transmitters have phase reversed to one another, their destructive combination hence results in a totally faded flat channel. It is even worse that, for a static/quasi-static receiver, this terrible situation may continue for a long time relative to the time interleaving length. Under such circumstances (flat and/or slow fading), a poor performance due to burst errors is expected for COFDM systems. Therefore, it is important to ‘create’ diversity for solving the problem without affecting the receiver design (i.e., backwards compatible). Cyclic delay diversity (CDD) [8], [9] is one of the promising diversity schemes for OFDM systems that artificially increases the frequency selectivity by transmitting plural versions of the same signal from single/multiple BS(s) with different cyclic delays. With similar concept as CDD, group-wise scrambling diversity (GSD) [10] increases the frequency selectivity by forming multiple subcarrier-groups, each rotated by a random phase independently for all BSs in an SFN. However, the composite channel response is thus divided into multiple groups with discontinuities in the frequency-dimension that may cause incorrect channel estimation via a typical approach of frequency-dimension interpolation. Moreover, both the CDD and GSD schemes provide no time diversity against the slow fading nature. In this paper, we propose a novel correlated scrambling diversity (CSD) scheme where the signal of each BS is independently scrambled, in both frequency- and timedimensions, before transmission. For each BS, the phase rotations of contiguous subcarriers are correlated for composite channel response being continuous in frequencydimension. On the other hand, the phases of contiguous symbols are rotated with controlled correlation depending on whether temporal interpolation is used for channel estimation or not. As a consequence, improved performance thanks to the diversity gain can be achieved with a simple channel estimation method by receivers.

978-1-4244-1722-3/08/$25.00 ©2008 IEEE. Authorized licensed use limited to: Industrial Technology Research Institute. Downloaded on December 12, 2008 at 03:56 from IEEE Xplore. Restrictions apply.

1

The rest of this paper is organized as follows. In Section II, the system model is presented. In Section III, GSD is briefly introduced and followed by a presentation of the proposed CSD scheme. A typical linear interpolation scheme for channel estimation is also introduced in Section IV as a supplement to the simulation results provided in Section V. Finally, we make some conclusions in Section VI.

where H comp denotes the composite channel response of the M M

A typical SFN configuration is shown in Fig. 1. There is one transmitter at the center of each cell. Each receiver may receive line-of-sight (LOS) signals as well as echoes from the neighboring transmitters. Fig. 2 shows the system model. It’s assumed that M BSs transmit the same signal, OFDM symbols, through different channels which are then received and combined at the receiving antenna of a UE. Under a further assumption that length of cyclic prefix (CP) of OFDM symbols is sufficiently large, the system model can thus be derived as follows.

Y = ∑ Hm ⊗ X + W ,

additive Gaussian noise, and ⊗ means element-wise product between vectors, A T means the transpose of a vector A. Eq. (1) can also be described as follows. (2) Y = H comp ⊗ X + W , individual channels as:

II. SYSTEM MODEL

M

signal to be transmitted, W = [W1 , W2 , …, WN ]T denotes an

(1)

m =1

where Y = [Y1 , Y2 ,… , YN ]T is the received signal in N subcarriers of one OFDM symbol after N-point FFT operation, T H m = [H m1 , H m 2 , … , H mN ] includes N sub-channels from the

m-th BS to the UE, X = [ X 1 , X 2 ,…, X N ]T is the encoded

H comp = ∑ H m .

(3)

m =1

At the cell edge between transmitters, it may happen that the composite channel given by Eq. (3) becomes a totally faded flat channel due to tiny delay spread and destructive combination of various signal paths. A worse case is that, for a static/quasi-static receiver, this terrible situation may continue for a long time. III. DIVERSITY SCHEMES This section presents two diversity schemes for improving the cell edge performance. The first one is the GSD scheme, and the other is the proposed CSD scheme. A. GSD Scheme The process of GSD is summarized in Fig. 3. In the scheme, the N subcarriers are divided into G groups, each containing N group subcarriers, that is, N = G ⋅ N group . Then, the phases of all subcarriers in the same group are rotated equally, while different groups are rotated independently. It can be described as follows. M

{

}

Y gsd = ∑ H m ⊗ X mgsd + W ,

(4)

X mgsd = D mgsd ⊗ X ,

(5)

m =1

where

D

gsd m

= [Pm ,1 , Pm , 2 ,… , Pm ,G ] , T

(6)

in which Pm, g is a 1× N group vector whose elements are all Fig. 1. SFN system configuration.

equal to

e

j⋅θ m , g

,

θ m, g

is the rotated phase for the g-th group of

the m-th BS. Rewriting Eq. (4) in composite form as in Eq. (2) as follows: (7) Y gsd = H gsd ⊗ X + W , comp

where

Fig. 2. System model.

Fig. 3. The process of GSD.

2 Authorized licensed use limited to: Industrial Technology Research Institute. Downloaded on December 12, 2008 at 03:56 from IEEE Xplore. Restrictions apply.

M

(

)

H gsd = ∑ H m ⊗ D mgsd . comp m =1

To clarify the characteristic of GSD, assume the individual channels form the m-th BS to the UE are flat faded. In such case, it can be easily shown that the composite channel given by Eq. (3) is also flat faded with a response depending on the constructive or destructive combination. Instead, for an SFN adopting the GSD scheme, the composite channel given by Eq. (8) can be further simplified as M

(

m =1

T

M

M

B. The Proposed CSD Scheme Fig. 4 shows the process of the proposed CSD scheme. Defining scrambling symbol as the value determining rotation of each subcarrier, the scrambling pattern is the way that scrambling symbols are organized along the time- and frequency-dimension. In the proposed scheme, the scrambling symbols are correlated for contiguous subcarriers (frequencydimension) and/or contiguous symbols (time-dimension). Corresponding to Fig. 4, the CSD can be described as follows: M

{

}

Y csd = ∑ H m ⊗ X csd +W, m

(10)

csd X csd m = Dm ⊗ X ,

(11)

m =1

where

[

D csd m = e

j ⋅θ m ,1

,e

j ⋅θ m , 2

,…, e

],

j ⋅θ m , N T

m =1

(

)

j⋅θ m , k

is the scrambling symbol for k-th subcarrier of m-th BS. Rewriting Eq. (10) in composite form as in Eq. (2) gives rise to

(14)

To clarify the characteristic of the proposed CSD scheme, assume again the individual channels form the m-th BS to the UE are flat faded. The composite channel given by Eq. (14) can be further simplified as M

m=1

(

) T

(15)

M ⎡M ⎤ j⋅θ j⋅θ = ⎢ ∑ H m ⋅ e m ,1 , … , ∑ H m ⋅ e m , N ⎥ . m =1 ⎣ m=1 ⎦

where

H m is a scalar representing channel fading factor

associated with m-th BS. Obviously, the composite channel is no longer flat but depends on the combination results whose feature is further determined by the scrambling pattern. There are many ways to generate an appropriate scrambling pattern for satisfying some criteria which will be discussed in the next section. One method is proposed here as an example. Let g be an Nx1 vector, and

⎧ 1 j⋅θ a ⎪ 2 ⋅ e , for n = 1, ⎪ j⋅θ ⎪1 ⋅ e b , for n = 1 + d , gn = ⎨ ⎪ 1 ⋅ e j⋅θ c , for n = 1 + 2d , ⎪2 ⎪0, otherwise, ⎩

(16)

g n denotes the n-th element of g, 2d is a parameter defined by the pattern generator, and θ i , i = {a, b, c} is a random phase ranging between ± π .

where

Let G be the N-point DFT of g. Then, the phases of the scrambling symbols can be chosen as

Θ = [θ1 ,θ 2 ,…,θ N ]

T

(

)

= G − G min ⋅

(12)

in which Dcsd is the scrambling pattern for m-th BS, and m

e

M

. H csd = ∑ H m ⊗ D csd m comp

(9)

⎡ ⎤ = ⎢∑ H m ⋅ Pm,1 , ∑ H m ⋅ Pm, 2 ,… , ∑ H m ⋅ Pm,G ⎥ . m =1 m =1 ⎣ m=1 ⎦ where H m is a scalar representing channel fading factor associated with the m-th BS. Obviously, the composite channel is no longer flat but a randomly weighted sum of the channels between the receiver and transmitters for each subcarrier group. Therefore, frequency diversity is introduced across subcarrier groups. However, discontinuities of composite channel responses inherent in the GSD scheme make channel estimation even more difficult (i.e., interpolation across subcarrier groups is forbidden). M

(13)

where

= ∑ H m ⋅ D csd H csd m comp

)

H gsd = ∑ H m ⋅ D mgsd comp

Y csd = H csd ⊗X+W , comp

(8)

2π , G MAX − G min

(17)

where G MAX and G min means the maximum and minimum value of

G , respectively. In this way, the scrambling

symbols expected to be strongly correlated for contiguous subcarriers can be obtained. No prior information about the scrambling patterns is needed by receivers that can estimate the composite channel given by Eq. (14) via a typical pilotbased linear interpolation approach (to be addressed in detail in Section IV).

Fig. 4. The process of CSD.

3 Authorized licensed use limited to: Industrial Technology Research Institute. Downloaded on December 12, 2008 at 03:56 from IEEE Xplore. Restrictions apply.

On the other hand, a specific Doppler frequency can also be applied to g so that Θ changes gradually symbol-by-symbol without inter-carrier interference (ICI) effect, and the time diversity is thus introduced to the system. IV. CHANNEL ESTIMATION Let us introduce a typical pilot-based linear interpolation approach for channel estimation as a supplement to the simulation results to be presented in the next section. Let X Pilot denote signals on a subset of subcarriers on which known pilot symbols are placed. Then, the corresponding received subcarriers in the receiver can be expressed as (18) YPilot = H Pilot ⊗ X Pilot + WPilot , where YPilot , H Pilot , and WPilot mean the received data, channel response and noise, respectively, on the same subset of subcarriers corresponding to the pilot subcarriers. Channel estimation of pilot subcarriers is simply as follows.

[

ˆ ˆ ˆ ˆ H Pilot = H p1 , H p 2 , … , H pP

]

T

=

YPilot , X Pilot

(19)

where Hˆ pi is i-th pilot subcarrier and division is performed element-wise. Then, channel estimates for data subcarriers

Parameters

Channel estimation Receiver location 2d (CSD) Number of groups regarding active subcarriers (GSD)

{ } ( {

}

{ })

{ } ( {

}

{ })

1024 600 15kHz 66.67μs 16.67μs 83.34μs 16-QAM Turbo code, rate 1/3 8 12 OFDM symbols 2 Single path or typical urban 6-paths (TU6) Linear interpolation Cell boundary 20 20

parameter 2d such that the coherent bandwidth of the composite channel is greater than the interval between two contiguous pilots. In other words, the parameter 2d is designed for the trade-off between diversity and channel estimation accuracy.

Hˆ pi and Hˆ p (i+1) can be obtained as

k ⎫ ⎧ Hˆ dik = ⎨Re Hˆ pi + Re Hˆ p (i +1) − Re Hˆ pi ⋅ ⎬ D + 1⎭ , ⎩ k ⎫ ⎧ + j ⎨Im Hˆ pi + Im Hˆ p (i +1) − Im Hˆ pi ⋅ ⎬ D + 1⎭ ⎩

Values

Number of subcarriers Number of active subcarriers Subcarrier spacing Useful symbol duration Guard interval Total symbol duration Modulation Channel coding Turbo decoder iteration Block interleaver size Number of cells Channel model for each cell

V. SIMULATION RESULTS (20)

where Hˆ dik is the channel estimate of k-th subcarrier between

Hˆ pi and Hˆ p ( i+1) , k = {1,2,..., D} , and D is number of data subcarriers between Hˆ pi and Hˆ p ( i +1) . The process is illustrated in Fig. 5. Recalling that CSD can increase frequency selectivity of the composite channel (refer to Eq. (14)). In order for accurate channel estimates through the linear interpolation approach, it’s expected that composite channel response is likely coherent between two contiguous pilot subcarriers. For the scrambling pattern generated with g n given by Eq. (16), the coherent bandwidth of the composite channel is inversely proportional to 2d and can be given approximately by N . Therefore, one should carefully design the 2π ⋅ (2d )

Computer simulation is performed to evaluate the proposed CSD scheme and the GSD scheme. The simulation platform is established according to the LTE standard [11]-[13], and the simulation parameters are described in Table 1. In this simulation, the performances of block error rate (BLER) versus the required signal to noise ratios (SNR) are evaluated. The composite channels by using GSD and CSD are shown in Fig. 6. Fig. 7 shows the BLER curves of CSD and GSD using ideal channel estimation, i.e., channel state information (CSI) is known at the UE. The channel model used is single-path model. Under the impractical assumption of known CSI, both schemes achieve similar performance, i.e., both schemes provide comparable diversity gain to the system. Fig. 8 shows the BLER curve of CSD and GSD using linear interpolation as the channel estimation scheme. It can bee seen that CSD outperforms GSD by 3.5dB for BLER=10-2 when imperfect channel estimation effect is taken into account. 4 3.5 Composite Channel Power

between

TABLE 1 SIMULATION PARAMETERS

3 2.5 2 1.5 1 CSD GSD

0.5 0

Fig. 5. Linear interpolation channel estimation scheme.

0

50

100

150 Subcarrier Index

200

250

300

Fig. 6. The composite channel response of CSD and GSD.

4 Authorized licensed use limited to: Industrial Technology Research Institute. Downloaded on December 12, 2008 at 03:56 from IEEE Xplore. Restrictions apply.

10

10

BLER curve with ideal CE over the 2-Single Path Channel

0

10

No Diversity GSD CSD

-1

Post-Turbo Block Error Rate

Post-Turbo Block Error Rate

10

-2

10

-3

10

-4

5

10

15

20

SNR (dB)

25

30

35

BLER curve with CE Methods over the 2-Single Path Channel 0

Post-Turbo Block Error Rate

10

10

10

No Diversity GSD CSD

-1

[3] [4]

-2

[6]

-3

[7] [8]

-4

5

10

15

20

No Diversity GSD CSD

-1

-2

10

-3

10

-4

5

7

9

11

13

15

17

Fig. 9. The BLER curve of CSD and GSD using linear interpolation channel estimation with TU6 channel model.

[5] 10

10

BLER curve with CE over the 2-TU6 Channel

SNR (dB)

Fig. 7. The BLER curve of CSD and GSD using ideal channel estimation with single path channel model.

10

10

0

25

30

35

SNR (dB)

[9]

Fig. 8. The BLER curve of CSD and GSD using linear interpolation channel estimation with single path channel model. [10]

For a more realistic channel model TU6 (typical urban 6paths), the corresponding results using the same channel estimation scheme are shown in Fig. 9. It can be seen that, for BLER=10-2, CSD outperforms GSD by 1.5dB, and exhibits more than 2dB gain compared to original system without any diversity scheme. VI. CONCLUSIONS In an OFDM-SFN system, UE at cell edges experiencing flat and/or slow fading cannot operate well. We have presented a novel correlated scrambling diversity scheme to overcome this problem. Through computer simulations, the proposed CSD scheme has been shown to provide significant diversity gain compared the original system. Moreover, it has also been shown that the CSD scheme outperforms the existing GSD scheme when the channel estimation effect is considered.

[11] [12]

[13]

ETSI 300 744 – V.1.5.1 “Digital Video Broadcasting (DVB): Framing structure, channel coding and modulation for digital terrestrial television”, June 2004. ETSI EN 302 304 – V1.1.1 “Digital video broadcasting (DVB); transmission system for hand-held terminals (DVB-H)”, Nov. 2004. ETSI EN 300 401 "Radio broadcasting systems; Digital Audio Broadcasting (DAB) to mobile, portable and fixed receivers", May 1997. Telecommunications Technology Association in Korea, TTAS.KO07.0024, “Radio Broadcasting Systems; VHF Digital Multimedia Broadcasting (DMB) to mobile, portable and fixed receivers,” October 2003. James K. Cavers, Mobile Channel Characteristics, Kluwer Academic Publishers, 2000. M. Bossert, A. Huebner, F. Schuehlein, H. Haas, and E. Costa, “On Cyclic Delay Diversity in OFDM Based Transmission Schemes”, in OFDM Workshop, 2002. Armin Dammann and Stefan Kaiser, “Performance of low complex antenna diversity techniques for mobile OFDM systems,” in Proceedings 3rd International Workshop on Multi-Carrier SpreadSpectrum & Related Topics (MC-SS’2001), Oberpfaffenhofen, Germany, Sep. 2001, pp. 53–64, ISBN 0-7923-7653-6. K. Akita, R. Sakata, and N. Deguchi, “Group-wise scrambling diversity for broadcast and multicast services in OFDM cellular system”, in Proc. IEEE Vehic. Technol. Conf. 2007-Fall (VTC’F07), Maryland, USA, pp.174–178, Sep. 2007. 3GPP TR 25.814 V7.1.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Physical Layer Aspects for Evolved UTRA (Release 7),” Oct. 2006. 3GPP TR 25.912 V7.2.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Feasibility study for evolved Universal Terrestrial Radio Access (UTRA) and Universal Terrestrial Radio Access Network (UTRAN) (Release 7),” Aug. 2007. 3GPP TR 25.913 V7.3.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Requirements for Evolved UTRA (E-UTRA) and Evolved UTRAN (E-UTRAN) (Release 7),” Mar. 2006

Manuscript received February 29, 2008. Corresponding author: Hsien-Wen Chang (e-mail: [email protected]; phone:+886-3-5915729; fax: +886-35820279).

REFERENCES [1] [2]

A. Mattsson, “Single frequency networks in DTV,” IEEE Trans. Broadcasting, vol. 51, no. 4, pp. 413–422, Dec. 2005. W.Y. Zou and Y. Wu, “COFDM: An Overview”, IEEE Trans. on broadcasting, Vo1.41, No.1, March 1995, pp 1-8.

5 Authorized licensed use limited to: Industrial Technology Research Institute. Downloaded on December 12, 2008 at 03:56 from IEEE Xplore. Restrictions apply.