ROBUST TIMING SYNCHRONIZATION FOR OFDM BASED WIRELESS LAN SYSTEM Sridhar Nandula, K Giridhar Telecommunications and Computer Networks Group (TENET) Department of Electrical Engineering Indian Institute of Technology Madras Chennai - 600 036, India +919884110242, +914422578386
[email protected],
[email protected]
Abstract— In this paper, a robust and efficient frame detection and symbol timing synchronization technique suitable for IEEE 802.11a wireless LAN system is proposed. The proposed method does frame detection using a threshold comparison mechanism and performs Orthogonal Frequency Division Multiplexing (OFDM) symbol boundary detection using correlation techniques. This algorithm is a novel combination of self and cross correlation information to achieve symbol timing synchronization. The proposed algorithm can robustly detect the symbol boundary even under low SNRs, high frequency offset, and multipath.
bit stream
bit stream
Channel Coding & Interleaving
BPSK/QAM Mapping
Decoding & deinterleaving
Chanl Equal
IFFT
FFT
Cyclic prefix & Windowing
Cyclic prefix Remover
RF Transmitter
Frame detection & Timing Synch
RF Receiver
Channel Estimation
Figure 1. OFDM Transmitter and Receiver
1. INTRODUCTION 2. OFDM-WLAN SYSTEM
In recent past, interest on high speed WLAN is increasing very rapidly. In 1999, LAN/MAN standards committee of IEEE came up with 802.11a (a new physical layer standard based on OFDM technology), which provides high speed data transmis sion with data rates from 6 to 54 Mbps. It operates in the 5 GHz unlicensed national information structure band.
Figure 1 shows a simplified transceiver structure of OFDM based 802.11a system. The OFDM signal can be generated by taking IDFT (Inverse Discrete Fourier Transform) of QAM or PSK symbols. As per IEEE 802.11a specification, each OFDM symbol consists of 48 data carriers, 4 pilot carriers and 12 null carriers. Hence, IDFT size is 64 point and can be implemented using efficient IFFT algorithm. The output of IFFT becomes one OFDM symbol, with duration of Ts (3.2µs). Each OFDM symbol is cyclically extended with 16 samples of duration Tg (0.8 µs) and they will be removed at the receiver. The cyclic prefix length should be more than the channel impulse response to avoid Inter Symbol Interference (ISI).
OFDM is a special form of multi-carrier modulation and is particularly suitable for transmission over dispersive channels. With the help of Gu ard Interval (GI) between the OFDM symbols, OFDM 1 WLAN system can combat impairments due to large multipath delay-spreads very effectively. OFDM symbols are cyclically extended and this cyclic prefix (CP) provides GI between adjacent OFDM symbols. Due to the CP, OFDM systems are also less sensitive to timing errors.
The expression for the OFDM baseband signal (output of IDFT) is:
1 = N
N / 2 −1
∑
j 2 π kn N
In this paper, we propose an algorithm for frame detection and symbol timing synchronization by exploiting the repetitive nature of the short preambles provided in the 802.11a preamble. Finding the symbol timing for OFDM systems is nothing but finding the beginning of the OFDM symbol. This can be achieved by finding any boundary in the preamble.
Where Ck is the QAM or PSK modulated complex signal and N = 64. This baseband signal is then up-converted, modulated to radio frequency (RF) and transmitted.
0-7803-7651-X/03/$17.00 © 2003 IEEE T his work was supported in part by Integrated SoftTech Solutions Pvt. Ltd., Chennai, India.
IEEE 802.11a is a packet-based communication system. Each packet is preceded by a preamble as defined in IEEE 802.11a specification. The preamble structure is as shown in figure 2.
xn
1
C ke
(1 )
k =− N / 2
This figure also shows the information regarding signal field and data payload. As shown in the figure, the preamble consists of 10 short symbols each having 0.8µs duration, and two long preambles of 3.2µs duration each. PLCP Preamble 10 short+2 long symbols
t1 t2
t3
t4
t5
SIGNAL field One OFDM Symbol
t6 t7
t8
t9
Where A(n) is the auto-correlation magnitude and l is the size of the moving average filter and it is chosen as 3. The falling edge of the curve corresponds to the (N-CP) th sample. We can detect this falling edge by observing the slope of the curve. However, at low SNRs and high delay spread situations, exact detection of this edge is difficult. This edge can be localized with the help of cross correlation of the received sequence. If we do cross correlation of the received sequence with the local copy of the short symbol, we get peaks at the end of each short symbol. However, the frequency offset of the local oscillator disturbs the magnitude of these cross correlation peaks significantly. Instead of averaging this cross correlation for one short symbol, if we average over more short symbols, as indicated by (4), we can still detect the peak.
DATA Variable Number of OFDM Symbols
t10
GI2
10 x 0.8 =8us Signal detect, AGC, Timing Synch, Coarse Frequency Estmation
T1
T2
2 x 0.8 + 2 x 3.2 =8us Fine Frequency Estmation, Channel Estimation
M
C ( n) = ∑
Figure 2. Frame format and Preamble structure of IEEE 802.11a
l= 0
3. DETECTION OF FRAME
N
*
∑ r (l * N + k + n)s
*
(l * N + k )
( 4)
k =1
Where s(n) is the local copy of the short symbol. M is the number of short symbols over which we are averaging the cross correlation. The simulation result shows that this type of averaging can with stand ±20ppm frequency offset, which is the worst-case possible frequency offset specified in 802.11a standard.
At receiver, the received signal is correlated with itself with a delay of one short symbol, given by
A( n) = ∑ r ( k + n) r ( k + n + L)
N
( 2)
k =0
6
Where r(n) is the received sequence, A(n) is the correlation output and L is the length of the short symbol. The incoming frame at receiver can be detected by comparing the magnitude of auto-correlation result with some threshold. It is advisable to have a dynamic threshold based on incoming signal power.
Cross Correlation magnitude Auto Correlation magnitude
5
Detect this boundary Magnitude
4
3
2
The initial 2-3 short symbols are assumed to be non-reliable, as Automatic Gain Control (AGC) logic requires some time to finalize the gain setting.
1
0 1
4. SYMBOL BOUNDARY DETECTION
Recevied Sample Index
In this paper, we propose a robust algorithm to detect OFDM symbol boundary using auto-correlation and cross correlation of short preamble. In (2) the value of N should be in between 16 and 144 and a multiple of 16. For any particular value of N, if we plot the auto-correlation magnitude values, we get a curve as shown in figure 3. The curve rises to some value, remains flat for about N-CP samples duration and then falls down as shown. In our algorithm, we detect the index of the (N-CP+1)th sample when counted from the start of the preamble.
Figure 3. Auto-Correlation curve and Cross-Correlation peaks for ideal case (No noise, no multipath and no frequency offset) Figure 3 shows the auto-correlation curve and the cross correlation peaks in the case of an ideal channel. Our objective is to detect the cross correlation peak at which the falling edge of the auto-correlation curve starts. This can be achieved by tracking the slope of the curve with the help of another dynamically set threshold. To detect the corresponding cross correlation peak, we need to do peak search by taking a window of 16 cross correlation magnitude values around that falling edge. The window size can be decided through simulations. The detected peak corresponds to the index of the (N-CP+1)th sample when counted from the start of the preamble. Thus, in the
The auto-correlation magnitude values are passed through a moving average filter to smoothen the curve. The moving average filter is defined by
Y ( n) =
l 1 ∑ A(n + k ) + A(n ) 2l + 1 k =− l
10 19 28 37 46 55 6 4 7 3 8 2 9 1 100 109 118 127 136 145 154 163 172 181 190
(3) 2
proposed algorithm, cross correlation is used to localize the exact boundary of the short symbol.
Figure 5 shows the auto-correlation magnitude curve and cross correlation magnitude peaks at 10dB SNR, with an RMS multipath delay spread of 60ns and with zero frequency offset.
The difference between the detected boundary and the actual boundary is the boundary detection error. This boundary detection error results in corresponding rotation of the signal constellation in frequency domain. This rotation can be taken care by channel estimation and equalization up to some extent.
Figure 6 shows the same auto-correlation and cross correlation magnitude plot for 7dB SNR, 100ns rms delay spread and 200 kHz frequency offset. This is one of the worst-case channels that we could expect in indoor wireless LAN. In Figure 6, the effect of frequency offset on the cross correlation peaks can be seen clearly. The magnitude of the cross correlation peaks is drastically reduced compared to the auto correlation magnitude. In this case also, even with very low magnitude peaks we are still able to identify the peaks and detect the symbol boundary. This is due to the extra averaging that is used in cross correlation (i.e., over more number of short symbols ) as shown in (4).
5. SIMULATION RESULTS We built transmitter and receiver prototypes based on 802.11a standard and tested OFDM implementation in the presence of AWGN noise, exponential multipath channel and local oscillator frequency offset. Multipath Channel Model 0.6
7
0.5
Auto Correlation magnitude Cross Correlation magnitude
6 5
Detect this boundary
0.3
Magnitude
Power
0.4
0.2 0.1
4 3 2
0 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
Tap number
0 1
Figure 4. Exponential channel Model considered for 60 ns rms delay spread
19
28
37
46
55
64
73
82
91 100 109 118 127 136 145 154 163 172 181 190
Received Sample Index
Figure 6. Auto-Correlation curve and Cross-Correlation peaks for SNR = 7dB, multipath delay spread = 100 ns and frequency offset = 200 kHz
Figure 4 shows the exponential channel model considered for 60 ns rms delay spread. The maximum rms delay spread that we can expect in indoor wireless channels is 60-100ns [1]. The exact value depends on the structure and fittings of the building.
Selection of various parameters in the equations is based on iterative simulations. To have better nois e averaging, N is chosen as 80 in (2). In this case, the duration of the flat portion of auto-correlation magnitude curve will be for 64 samples and we detect the index of 65th sample. In equation 4, value of M is chosen as 5 and the value of N is chosen as 16.
6
Auto Correlation magnitude Cross Correlation magnitude 5
Detect this boundary 4
Magnitude
10
3
Figure 7 shows the simulation results. It shows the synchronization error rates obtained for different types of channel models considered.
2
1
0 1
10 19 28 37 46 55 64 73 82 91 100 109 118 127 136 145 154 163 172 181 190
Received Sample Index
Figure 5. Auto-Correlation curve and Cross-Correlation peaks for SNR=10dB and multipath delay spread 60 ns and no frequency offset 3
Consumer Electronics, vol. 44, pp. 217-225, February 1998. [5] Chia-Sheng Peng, Chian-Hung Cho, and Kuei-Ann Wen, “Frame Synchronization and Digital AGC for OFDM based LAN,” Institute of Electronics, National Chiao Tung University, Taiwan.
Synchronization Error Rate Curves
0
10
Freq offset 200kHz Multipath Trms=60ns AWGN Multipath Trms=60ns & freq offset 200kHz
-1
Synchronization Error Rate
10
-2
10
-3
10
-4
10
2
4
6
8 SNR in dB
10
12
14
Figure 7. Synchronization Error Rate results for AWGN, Multipath (Trms= 60ns), and Frequency offset (200 kHz) models
6. CONCLUSIONS Robust frame detection and symbol timing detection algorithms were proposed for IEEE 802.11a WLAN system. In particular, a novel combination of auto-correlation and cross-correlation information was used to increase the reliability of the frame boundary detection algorithm. These algorithms were simulated for AWGN, exponential multipath channel and frequency offset models and some of these results were presented. Satisfactory synchronization performance was obtained to meet the Packet Error Rate (PER) performance criterion specified in 802.11a specification.
7. ACKNOWLEDGEMENTS The authors thank Mr. K.V. Srinivas, Ph.D. Scholar, IIT Madras for useful discussions.
8. REFERENCES [1] Richard van Nee and Ramjee Prasad. OFDM for Wireless Multimedia Communications, Boston: Artech House, 2000. [2] Timothy M. Schmidl and Donald C. Cox, “Robust Frequency and Timing Synchronization for OFDM,” IEEE Trans. on Communications, vol. 45, no 12, pp. 1613-1621, December 1997. [3] IEEE Std. 802.11a-1999, “Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: High-speed physical layer in 5GHz Band,” 1999 edition. [4] M. H. Hsieh and C. H. Wei, “Channel estimation for OFDM systems based on comb -type pilot arrangement in frequency selective fading channels,” IEEE Trans. 4
