Joint Estimation of Symbol Timing and Carrier Frequency Offset Over A Fast Time-Varying Multipath Channel

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 8

Joint estimation of symbol timing and

Carrier frequency offset over a


Fast time-varying multipath channel
ABSTRACT:
1. In this paper, we present a novel joint algorithm to estimate
the symbol timing and carrier frequency offsets of wireless orthogonal
frequency division multiplying (OFDM) signals.

2. We develop a general ML estimation algorithm that can


accurately calculate symbol timing and carrier frequency offsets over a
fast time varying multipath channel.

3. To reduce overall estimation complexity, the proposed


scheme consists of two estimation stages (i) Coarse synchronization –
LS method and (ii) Fine synchronization – MS method.

4. This coarse-to-find method provides good tradeoff between


estimation accuracy and computational complexity.

INTRODUCTION:
1. OFDM technique is used for providing high speed data
transmission.
2. Because of Intersymbol interference(ISI) and Intercarrier
interference(ICI) synchronization errors can result.
3. To avoid this error we go to identify the begining of the
individual orthogonal frequency division multiplexing(OFDM)
symbols.
4. We can find the offsets jointly or separately.
5. OFDM is used in mobile multipath environments, a fast time-
varying multipath channel model is more desirable than a
White Gaussion Noise channel model for simulating real
wireless OFDM communication system.
6. This estimation involves a two-step process (i) Coarse
synchronization – to find rough symbol offsets (ii) Fine
synchronization – to find fair offsets based on the coarse
synchronization.
7. The proposed fine synchronization technique can obtain more
accurate results.
8. While its computational complexity is drastically reduced due
to the two-step estimation process.
9. The result of this paper is organized as follows:
(i) We introduce the OFDM signal model over time-

Varying multipath channel in section-II.

(ii) The coarse synchronization technique is then


explained in details in section-III, followed by the fine
synchronization technique in section-IV.

10. Simulation results are provided in section-V. Finally we


conclude our paper in section-VI.

OFDM SIGNALS AND CHANNEL MODEL:


1. In an OFDM system, N complex data symbols are modulated
onto N-subcarriers by using the inverse fast fourier transform(IFFT) on
the transmitter side.

2. The last Nc IFFT samples are used to form a guard interval(GI)


that is inserted at the beginning of each OFDM symbol.
3. The baseband modulated signal s(n), after the parallel to serial
convertsion and IFFT can be expressed as, Equation ->(1)

Where dk(i) – data symbols used in modulating the k-th subcarrier of


the i-th frame (or) data block.

Frame period is M=N+Nc.

4. These data symbols, coming from a certain modulation


constellation in a finite symbol set, can be assumed to be
equiprobable and statistically independent of different subcarriers and
frames.

5. dk(i) can therefore be approximated as zero mean random


variables with correlation

Equation

Where * is a conjugate operator. g(.) is a dicrete rectangular window


function of size M.

X(n) is a output signal after s(n) has passed through a fast time-
varying multipath channel with impulse response h(n,l)

6. In wireless communication receivers are subjected to reflection


and scattering from natural and man-made objects.

7. Received signals after having been sampled by an analog to


digital converter(ADC) can be expressed as,

Equation

Where ne belongs to [0,M-1] is the symbol timing offset.


8. The timing offset has an unkown integer value (ie) an
integermultiple of the sampling period and the fractional timing offset
can be be neglected in OFDM systems because it has little affect on
OFDM performance.

9. ‘ ‘ denotes the carrier frequency offset caused by the Doppler


effect and the instability of carrier frequency oscillators at both the
transmitter and the received ends.

10. w(n) is white complex gaussion noise zero mean and a


variance of ‘ ‘ , which is independent of the signal dk(i).

11. The autocorrelation function of a received OFDM signal r(n)


is given by

Equation

12. The autocorrelation function Rr(n,m) is M-periodic in ‘n’ for


every m, or Rr(n,m)=Rr(n+M,m).

13. The received signal r(n) is cyclostationary but is not a wide


sense stationary process (wssp).

14. We can rewrite Rr(n,m) as compact matrix format when


m=N.

Equation

A – Toeplitz matrix of size M*(M+Nc-1).

By taking the modulus operation of Rr, b= equation

Where b->

The Least Square method(LS) estimator of ne is given by

Equation
Where b^-> estimate of b.

||.||2 denote the Euclidean normal.

Since Rr*(n,m)=Rr(n,-m) in Rr, we obtain the frequency offset ‘ ‘


estimation as

Equation

FINE SYNCHRONIZATION:
1. We will discuss the fine synchronization techniquefor timing
and frequency offsets estimation based on the ML function.

2. Received signals r(n) are independent of each other expect of


signals that are spaced N samples apart and that are correlative within
the same OFDM symbol interval, due to the GI.

3. r(n) can be modeled as an approximate complex Gaussian


process using the central limit theorem, provided that N is large
enough.

4. We assume an observation window to be a vector r=[r(0),


……..r(2N+Nc-1)] composed of 2N+Nc consecutive samples of r(n). The
length of this window has been selected sothat there are at least Nc
correlated samples, regardless of timing offset ne.

5. The autocorrelation function is not zero when n is within a


limited region and m=N. We can rewrite the autocorrelation function
for the observation vector r as

Equation

Where

6. To derive the ML estimation of symbol timing and carrier


frequency offsets over a fast time-varying multipath channel as was
previously defined, we can classify the ML estimation into three cases
depending on the timing offset ne as follows.

CASE 1:

0<=ne<=N-Nc
1. The relative positioning of OFDM symbols and an observation
window in case one is shown in fig1. To obtain the joint ML estimation
of ne and ‘ ‘ we calculate a likelihood function.

2. Under the previous assumption, the correlation between is


nonzero only in the regions F1(1) and F2(1), where F1(1)={n|
ne<=n<=ne+Nc-1} and F2(1)={n|ne+Nc<=n<=ne+2Nc-2}.

3. The additional region F2’ is the multipath spread of the GI of


the k-th OFDM symbol.

4. The joint probability density function (pdf) of vector r can be


expressed as

Equation

Where R=F(1)UF2(1), p(.) denotes the PDF of the variables.

By maximizing the log-likelihood function we get join ML


estimators of ne and ‘ ‘

Ne^

CASE 2:

N-Nc+1<=ne<=N-1
Fig2 displays the relative positioning of OFDM symbols and an
observation window in case two . The correlation of sampling signals
within the observation window is nonzero only in regions.
F1(2), F2(2) and F3(2), where

F1(2)={n|ne<=n<=ne+Nc-1},

F2(2)={n|ne+Nc<=n<=N+Nc-1},

And F3(2)={n|0<=n<=ne+Nc-N-2}.

The region F2(2) is the multipath spread of the GI of the k-th OFDM
symbol. Case two differs from case one in that the length of region
F2(2) is reduced due to limited observation window length.

The region F3(2) is the multipath spread of the GI of the (k-1)-


th OFDM symbol.

The joint PDF of vector r in this case has the same formula as
(I) expect that R=F1(2)UF2(2)UF3(2).

By maximizing its log-likelihood function the joint ML


estimators of ne and ‘ ‘ can be obtained as

Equation

CASE 3:
N<=ne<=N+Nc-1
The relative positioning of OFDM symbols and an observation
window in case three is represented in Fig3.

Unlike both case one and case two, the observation window is
continuously shifted to the left, which contains F1(3), F2(3) and F3(3).

Where F1(3)={n|ne<=n<=N+Nc-1},

F2(3)={n|0<=n<=ne-N-1} and

F3(3)={n|ne-N<=n<=ne-N+Nc-2}.
F1(3) is part of the GI in the k-th OFDM symbol. F3(3) similar to F3(2) is
the multipath spread of the GI of the (k-1)th OFDM symbol.

The joint PDF of vector r in this case is the same as that in (I)
expect that R=F1(3) U F2(3) U F3(3).

By maximizing the log-likelihood function the joint ML


estimators of ne and ‘ ‘ can be obtained.

From the previous discussion we can conclude that the


estimators ne^ and ‘ ‘ are tightly coupled with each other.

The estimation procedure can be described as follows. Ne


takes a value from 0 to N+Nc-1 and ‘ ‘ is then computed for each
given ne^ from 0 to N-Nc-1 and ‘ ‘ is then computed for each given ne
accordingly to (b),(d) and (f).

Ne^ is computed based on (a),(c) &(e). The expected ‘ ‘ is


finally obtained from (b),(d) and (f) after ne has been estimated.

You might also like