* Address correspondence to K.H. Chang (khchang@inha.ac.kr)
* This work was supported by grant No. R01-2003-000-10685-0 from the Basic Research Program of the Korea Science & Engineering Foundation.
Link-level Performance of Satellite DMB System A
considering Frequency & Timing Synchronization Error and HPA Non-linearity
SungHo Park
The Graduate School of Information Technology & Telecommunications
Inha Univ.
Incheon, Korea yannyplus@empal.com
InSuk Cha
The Graduate School of Information Technology & Telecommunications
Inha Univ.
Incheon, Korea inhaucha@hanmail.net
KyungHi Chang The Graduate School of Information
Technology & Telecommunications Inha Univ.
Incheon, Korea khchang@inha.ac.kr Abstract —The DAB (Digital Audio Broadcasting) system
which is based on Eureka-147 employing COFDM (Coded Orthogonal Frequency Division Multiplexing) technique is evolved into DMB (Digital Multimedia Broadcasting) system that is categorized into Terrestrial DMB and Satellite DMB.
The Satellite DMB is multi-channel multimedia broadcasting service for the reception by the portable or vehicular receiver.
Link level performance of Satellite DMB System A considering the sensitivity of HPA non-linearity has been analyzed in this paper. We choose the appropriate value of HPA back-off to reflect major practical RF impairment. We also investigated time and frequency synchronization techniques, which were suitable for Satellite DMB system A. Finally, we verified the overall link level performance considering the above RF impairment and frequency offset.
Keywords – DMB, Satellite, OFDM, Nonlinearity, HPA, Synchronization
I. INTRODUCTION
The Satellite DMB System is designed to provide high-quality, multi-service digital multi-media broadcasting for reception by vehicular, portable and fixed receivers. Satellite DMB System is capable of offering variable multi-media quality up to high-quality multi-media comparable to that of obtaining from consumer digital recorded media. It can also offer various data services and different levels of conditional access and the capability of dynamically re-arranging the various services contained in the multiplex [1].
At the World Administrative Radio Conference (WARC)-92, the 2.6 GHz frequency band was allocated to a group of East Asian countries including Korea. The DMB frequency spectrum is jointly managed with Japan on 2630 ~ 2655 MHz (25 MHz for SKT) and 2605 ~ 2630 MHz (25 MHz for KT).
Frequency allocation including Satellite DMB in Korea is shown in Fig. 1.
Figure 1. Frequency allocation including Satellite DMB in Korea.
Specification of the System E is adopted for the upper band
of DMB (2630 ~ 2655 MHz) in Korea. For lower band deployment, System A is the one of strong candidates for the time being.
In general, the satellite broadcasting systems amplify the signal to maintain high transmitted power using HPA (High Power Amplifier) on the Gap-filler or satellite. However, when the system using non-linear HPA has high input signal power, the signal contains non-linearity. The effect of non-linear HPA is appeared as the distortion of amplitude and phase. It is observed as AM/AM and AM/PM characteristics.
These characteristics attenuate the amplitude of transmitted signals, and cause the interference among adjacent cells by expanding bandwidth. By the result, overall system performance is degraded. These problems are more sensitive in the case of multi-carrier systems such as OFDM system [2].
In this paper, we analyzed the link level performance of the Satellite DMB System A considering HPA non-linearity, and we chose appropriate back-off value from the simulation results. We also investigated the suitable techniques of time and frequency synchronization for Satellite DMB System A.
This paper is organized as follows. In Section II, epitome of the Satellite DMB System A is introduced. In Section III, the HPA models are analyzed and the performances with varying back-off values are described. In Section IV, suitable time and frequency synchronization techniques for the Satellite DMB System A are investigated and the simulation results for traffic channel with synchronization error are presented. And conclusions are given in Section V.
II. SATELLITE DMBSYSTEM A 2.1. System Configurations
The block diagram of transmitter and receiver for the Satellite DMB System A is shown in Fig. 2. The Satellite DMB System A is COFDM system, and / 4π -DQPSK modulation is used. For channel coding, the system uses concatenated code which is composed of the RS code (204, 188, t=8) as an inner code and convolution code (k=1, n=4, K=7, average coding rate =1/2, mother code=1/4) as an outer code. Therefore, the Satellite DMB System A guarantees high error correction ratio and provides high quality of video and audio service. By using both time and frequency interleaver, the system becomes robust over selective fading channel.
The Satellite DMB System A supports two types of protection modes which are EEP (Equal Error Protection) and UEP (Unequal Error Protection), and provides four types of transmission mode based on Eureka-147. Therefore, the system provides flexible protection mode according to the importance of data, and adjusts transmission rate.
In this paper, our simulation was performed in the lower DMB band which is 2.605 ~ 2.630 GHz in Korea satellite band, and the simulation is considered only transmission mode III. Table 1 shows simulation parameters for transmission mode III of the Satellite DMB System A [1].
TABLE 1. SATELLITE DMBSYSTEM APARAMETERS FOR TRANSMISSION MODE III.
Transmission Mode III
Service Satellite / Cable / Terrestrial Available Frequency Lower than 3 ㎓
Number of Subcarrier 192
Subcarrier Spacing 8 ㎑
Traffic Symbol 125 ㎲
Guard Interval 31 ㎲
Total Symbol Duration 156 ㎲ Null Symbol Duration / Frame 168 ㎲
Coverage 24 ㎞ (Urban)
Frame Length 24 ㎳
Number of Symbol / Frame 153 ( 1 + 8 + 144)
Figure 3. Frame structure for Satellite DMB System A (transmission mode III).
A configuration of the transmission frame is different according to the transmission mode. The frame structure of
the mode III is shown in Fig. 3. SC (Synchronization Channel) is composed of Null Symbol and PRS (Phase Reference Symbol) for frame synchronization, carrier frequency synchronization, and differential-detection, etc.
FIC (Fast Information Channel) carries control information for MSC (Main Service Channel), and emergency data and transmission of emergency data. Therefore FIC doesn’t use time-interleaving causing time delay, and fixed coding rate (R=1/3) of UEP type is applied. MSC is consists of numerous sub-channels including useful traffic data, so it uses both UEP and EEP type of protect.
2.2 Simulation Conditions
In this simulation, we adopted channel model of ITU-R M.
1225. In indoor environment, the simulation was performed in Indoor B (3 km/h) channel. For outdoor environment, Pedestrian A (3 km/h), Vehicular A (60 km/h) and Vehicular B (120 km/h) channel was simulated. For each simulation, 1/2 convolutional coding was assumed.
III. LINK-LEVEL PERFORMANCE OF SATTLITE DMB SYSTEM A CONSIDERING HPANON-LINEARITY The Satellite DMB System A employs OFDM (Orthogonal Frequency Division Multiplexing) technique as a signal multiplexing method. OFDM system which has orthogonality among sub-carriers is robust for multipah fading, and provides high spectral efficiency. On the other hand, The OFDM system with numerous sub-carriers has high PAPR (Peak to Average Power Ratio) [4]. So, the OFDM system employing HPA, it’s performance is degraded due to the HPA non-linearity. The Satellite DMB System A uses back-off as one of the methods to compensate this problem. Consequently, the signal is processed in the linear region by applying appropriate back-off value. In this paper, IBO (Input Back-Off) for the power level of input signal was applied.
3.1 TWTA (Traveling Wave Tube Amplifier)
TWTA, which is used on the satellite, provides high power, but it has poor efficiency and reliability. In this paper, the simulation was based on the Saleh’s model. The AM/AM and AM/PM characteristics are as follows [5];
2
[ ( )] ( )
(1 a( ( )) ) A r t vr t
β r t
= + (1) Fiqure 2. Block diagram of transmitter & receiver for Sattelite DMB System A.
2 2
( ( )) [ ( )]
(1 ( ( )) ) r t r t
r t
φ φ
α
Φ = β
+ (2) Here, we set parameters asv=1, βa=0.25in Eq. (1), and
φ /12
α =π , βφ=0.25 in Eq. (2).
A [ ] ⋅
andΦ ⋅ [ ]
denoteamplitude and phase of the signal.
r t ( )
is an input signal. As increasing the amplitude of input signal, the amplitude and phase of output signal becomes more and more non-linear as shown in Fig. 4. Especially, phase degradation for sub-carriers is caused due to the phase non-linearity. By the result, ICI (Inter-Carrier-Interference) is induced. Therefore, linearity should be recovered by setting back-off input signal level below the saturation point.Figure 4. AM/AM & AM/PM characteristic curves of the TWTA.
Fig. 5 shows the simulation results which were performed under satellite B channel (60 km/h) with various IBO values from 0 dB to 15 dB. There is about 2.7 dB of degradation in the case of 15 dB IBO compared with the case of ideal linear amplifier.
Figure 5. CNR vs. BER performance varying IBO values under the Satellite B channel (TWTA).
3.2 SSPA (Solid State Power Amplifier)
The SSPA is used for the Gap-filler on the ground for the application of low power amplification and low cost.
The AM/AM and AM/PM characteristics are as follows [5];
2 1/ 2
0
[ ( )] ( ) 1 ( )
p p
A r t r t
r t A
=
+
(3)
[ ( )] 0r t
Φ = (4) The characteristics of Eq. (3) and (4) are shown in Fig. 6.
For large value of p, the model convergences to a clipping amplifier that is perfectly linear until it reaches its maximum output level.A 0 denote the saturation point, so the linear region is below this value. In this simulation, A0was set as 1.
The phase attenuation is negligible for SSPA, therefore it is assumed as 1.
Figure 6. AM/AM characteristic curves of the SSPA.
The simulation results are shown Fig. 7 and Fig. 8, applying SSPA (p=1) for the Pedestrian A (3 km/h) and Vehicular A (60 km/h) channel. As shown in these figures, as IBO value is increased, linearity of the system is recovered. There is about 2 dB of degradation in the case of 15 dB IBO compared with the case of ideal linear amplifier.
Figure 7. CNR vs. BER performance varying IBO values in the Pedestrian A (3 km/h) channel (SSPA).
Figure 8. CNR vs. BER performance varying IBO values in the Vehicular A (60 km/h) channel (SSPA).
IV. SYNCHRONIZATION TECHNIQUES FOR SATELLITE DMBSYSTEM A
The synchronization error is caused by timing and frequency errors. The synchronization procedure of satellite DMB system A is accomplished in according to following steps; coarse frame synchronization, integral frequency offset estimation, fine frame synchronization, and fractional frequency estimation. The NULL Symbol and PRS are utilized for this purpose. The structure of transmission frame including synchronization channel is shown in Fig. 9. Null Symbol is composed of zeros, used for initial frame synchronization. Its length is 168 ㎲ (345 sub-carriers) for the transmission mode III. PRS itself is used for the time and frequency synchronization except fractional frequency offset estimation. The synchronization procedure of Satellite DMB System A is shown in Fig. 10.
In this paper, the effects of time and frequency offset have been analyzed under AWGN. If the received signal including Null Symbol is
r t ( )
, it can be expressed as following,1 2
1
( ) 1 N ( ) j f tk ( )
k
r t s k e w t
N
− π
=
=
∑
+ (5)where N is the number of sub-carriers,
s k ( )
is a transmitted symbol (k=0, …, N-1) in frequency domain,f
k(fk =k T/ ) is a center frequency of k-th sub-carrier, T is a period of the OFDM symbol, andw t ( )
is AWGN withσω2. First, coarse frame offset (τ
F ) is detected and synchronized for the received signal. And then, using correlation between received PRSy k ( )
and local PRSx k ( )
in receiver, estimation and synchronization are performed for integralfrequency offset (∆fˆi) and fine frame offset (
τ ˆ
) using cyclic prefix.Finally, by detecting and correcting of fractional frequency offset ( ˆ
ff
∆ ), the synchronization process is completed.
Figure 10. Synchronization procedure for Satellite DMB System A.
4.1. Time Synchronization
( )
r t
in Eq. (5) is sampled in the receiver periodically. If the sampling interval is t T N= n/ + (ττ
: frame offset), the received sample can be expressed as following;1 2
0
1 (2 )/ (2 ) /
0
( ) 1 ( ) ( )
1 ( ) ( )
N j n Tk
T N n
N j n T j nk N
n
r k s n e w k
N
s n e e w k
N
π τ
π τ π
− +
=
−
=
= +
= +
∑
∑
(6)
Demodulated symbols after FFT could be expressed as following;
1 2
0
ˆ( ) 1
N( )
j mkm
s k r k e
N
− π
−
=
= ∑ (7) The demodulated symbol in Eq. (7) is reformulated by applying Eq. (6),
2
ˆ( ) ( ) ( )
j k
s k s k e
Tu k
π τ
= +
(8) whereu k ( )
is noise component in frequency domain. In Eq.(8), the received signal is expressed to account the phase rotation by AWGN in frequency domain. The amount of rotation is decided by
τ
and sub-carrier index k.4.1.1. Initial Frame Synchronization
The coarse frame synchronization is the process that seeks for starting point of the received frame. It should be
synchronized within about 10 ~ 20 times of sampling rate.
The coarse frame offset could be detected using the fact that the energy of Null Symbol is 0. Therefore, we used the energy ratio of the Null Symbol and PRS. We set up two windows whose length is equal to Null Symbol or smaller than it. Then, we shifted the windows and searched a peak point. This point means the starting point of PRS [6].
Example of windowing for coarse frame synchronization is shown in Fig 11. The expression for the coarse frame synchronization is in Eq. (9). This procedure is performed per frame.
* *
0 0
( ) / ( )
max
F n n n n
n
y y y y
ω ω
ω λ ω λ λ λ
λ λ
τ + + + + + +
= =
=
∑ ∑
(9)Here,
y
n is received signal,ω
is window size which should be less than the size of Null Symbol. This technique has disadvantage when CNR is low due to the rarely detected peak point. Especially, it is weak for the case of multipath channel. It could be improved by adjusting of window size.The simulation result is in Fig. 12 and it is normalized for the starting point of PRS.
4.1.2. Fine Frame Synchronization
The fine frame synchronization is the searching process to fine the accurate starting point of FFT. It prevents from ISI (Inter Symbol Interference) due to symbol timing offset. In this simulation, we used the technique which is using channel impulse response of the PRS [7]. In Eq. (10), the output of IFFT appeared as a form of channel impulse response with the reference of integral frequency offset and the maximum IFFT output of the correlation between the received PRS and the local PRS became fine frame synchronization timing. That is accurate integral frequency offset estimation should be preceded. The simulation result with AWGN (CNR=5 dB) is shown in Fig. 13.
{
* ˆ}
ˆ ( )
max
n IFFT Y k X k( fi N)τ= − − ∆ (10)
Figure 11. Example of windowing for coarse frame synchronization (transmission mode III).
Figure 12. Energy rate between two windows (Vehicular A channel (60 km/h), CNR=5 dB)
Figure 13. Channel impulse response of the phase reference symbol with AWGN (CNR=5 dB).
4.2. Frequency Synchronization
Generally, the carrier frequency offset is caused by two reasons. The first one is caused the misadjustment of LO frequency between the transmitter and the receiver. Another is due to the phase error of carrier frequency. In the Satellite DMB System A which uses DQPSK modulation, the phase error is appeared as frequency error. So we just performed the frequency offset estimation which is divided into integral frequency offset estimation and fractional frequency offset estimation.
4.2.1. Integral Frequency Offset Estimation
The unit of the frequency domain synchronization in the OFDM systems is a multiple of subcarrier spacing and it is expressed as BW/N (BW : channel bandwidth, N : the number of subcarriers). Therefore integral frequency offset is estimated with the accurate subcarrier spacing (i.e. The Satellite DMB System A : 8 kHz) . In this paper, we employed the technique using channel impulse response.
The output of IFFT is maximized at the position of symbol timing offset. As Eq. (11), it is detected maximum point of the IFFT output while saved PRS in the receiver is cyclic shifted as d. If the integral frequency offset is equal to d, the output of IFFT is appeared as channel impulse response at the symbol timing offset.
ˆi
max max
( ) *( )Nd amp
f IFFT Y k X k d
∆ = −
(11) Here, ˆ∆ is estimated integral frequency offset, fi X k( ) is local PRS in the receiver, Y k( ) is received PRS, and d is length of cyclic shift.
This technique could be used in spite of remaining symbol timing offset, but very complex. Fig. 14 shows that the simulation result under Vehicular A (60 km/h) channel for offset 16.
Figure 14. Channel impulse response of the phase reference symbol with Vehicular A channel (60 km/h). ( ˆ
fi
∆ =16)
4.2.2. Fractional Frequency Offset Estimation
The fractional frequency synchronization compensates for fractional carrier frequency offset and it is keeping into a small value within 1% of subcarrier spacing. In this simulation, the fractional frequency offset is estimated by GIB (Guard Interval Based) algorithm which utilizes cyclic prefix. If fractional frequency offset is not equal to 0, the phase between cyclic prefix and original part of that is different. So the correlation value includes imaginary part.
The imaginary part contains the information related to the fractional frequency offset. When the fractional frequency offset is less than half of subcarrier spacing, it can be compensated with the GIB algorithm. Fig. 19 shows the range of available detection range. The expression of estimated fractional frequency offset with the GIB algorithm is as following.
1 *
0
ˆ 1 arg 2
L
f n N n
n
f y y
π
−
= +
∆ =
∑
(12)Figure 15. The rage of fractional frequency offset estimation with multipath channel (vehicular A (60 km/h).
4.3. Link Level Performance considering Synchronization In this paper, we performed traffic channel simulation with frequency offset, and compensated it using GIB algorithm. The simulation result under Vehicular A channel (60 km/h) is shown in Fig. 16. The performance degradation of about 1 dB is occurred when the normalized frequency offset of 0.1 at the target BER 10-4. However, it can be compensated using the above mentioned algorithm.
Figure 16. CNR vs. BER performance for fractional frequency offset ( ˆ∆ =0.1) in the Vehicular A channel (60 km/h). ff
V. CONCLUSIONS
In this paper, we analyzed the overall link level performance of the Satellite DMB System A which is suggested as one of the Satellite DMB systems by considering the effect of HPA non-linearity. We verified the performance improvement when IBO technique is used for recovery of HPA linearity on Gap-filler. As this result, we confirm that the appropriate IBO values are ranged from 9 to 15 dB. Also, we investigated the effects of time and frequency offset in the Satellite DMB System A, and the techniques of estimation and compensation. Finally, we compared the link level performance with compensated frequency offset applying the above-mentioned techniques.
REFERENCES
[1] DSB Handbook, "Terrestrial and satellite digital sound broadcasting to vehicular, portable and fixed receiver in the VHF/UHF band," ITU-R, 2002.
[2] S. Andrecoli, H.G. McClureet, P. Banelli, and S. Cacopardi,
"Digital Linearizer for RF amplifiers," IEEE Trans.
Broadcasting, vol.43, no. 1, pp. 12-18, Mar. 1997.
[3] Recommendation ITU-R M.1225, Guidelines for Evaluation of Radio Transmission Technologies for IMT-2000, 1997.
[4] J.A.C. Bingham, "Multicarrier modulation for data transmission: An idea whose time has come," IEEE Commun. Magazine, Vol.28, No.5, pp.5-15, May 1990.
[5] Li Jungqiang, Ma zhengxin, Cao zhingang,
"Compensation of nonlinear distortion using a new method combining predistortion with reconstruction in OFDM systems," in Proc. ICCT, 2000.
[6] Chorng-Ren Seu, Yung-Liang Huang and Chia-Chi Huang, "Joint Symbol, Frame, and Carrier Synchronization for Eureka-147," in Proc. ICUPC, pp.693-697, Oct. 1997.
[7] F. van de Laar, N. Philips and J. Huisken, "Towards the Next Generation of DAB Receivers," EBU Technical Review, pp.46-59, Summer 1997.
[8] J.-J. van de Beek, M. Sandle and P.O. Borjesson, "ML estimation of timing and frequency offset in Multicarrier systems," Research Report TULEA 1996.
09, Division of Signal Processing, Lulea University of Technology, 1996.
