Carrier frequency offset (CFO) estimator with low complexity and high accuracy for orthogonal-frequency-division-multiplexing (CO-OFDM) systems

1. A method for carrier frequency offset (CFO) estimation for coherent optical orthogonal-frequency-division-multiplexing (CO-OFDM) broadband systems, comprising:

obtaining an initial estimate of a normalized CFO with an estimation range equal to +L/2 subcarrier subspacing using a single training symbol with L identical portions;

storing the initial estimate in a memory device; and

obtaining a maximum likelihood (ML) estimate of the CFO by performing a local grid search based on the initial estimate,

wherein the ML estimate is calculated by maximizing over a set of grid search results that fall within a range around the initial estimate defined by an empirically designed parameter, and

wherein the ML estimate is calculated as:

ε

ML

=

arg

⁢

⁢

max



ε

-

ε

^

ini



≤

ε

b

⁢

{

⁢

(

∑

l

=

1

L

-

1

⁢

p

⁡

(

l

)

⁢

ⅇ

-

j

⁢

⁢

2

⁢

⁢

π

⁢

⁢

ε

⁢

⁢

l

/

L

)

}

,

where ε is a normalized CFO, {circumflex over (ε)} ini is an initial CFO estimate ε b is an empirically designed parameter, and p(l) provides correlations.

2. The method of claim 1 , wherein the initial estimate is obtained without considering channel information and without a grid search.

3. The method of claim 1 , wherein the initial estimate is obtained by:

calculating averaged correlation coefficients between time-domain received symbols;

calculating a phase difference between any two consecutive ones of the averaged correlation coefficients;

calculating a weight factor of the phase difference; and

averaging a product of the weight factor and the phase difference.

4. The method of claim 1 , wherein the local grid search is limited to searching only within a pre-specified vicinity of the initial estimate.

5. The method of claim 1 , wherein the local grid search comprises a plurality of grid searches, and a number of the plurality of grid searches performed to obtain the ML estimate is determined by the empirically designed parameter and a grid interval.

6. The method of claim 1 , wherein the ML estimate is obtained by jointly estimating the CFO and an objective function, the objective function for performing the grid search to identify a plurality of candidate ML estimates.

7. The method of claim 6 , wherein the ML estimate is obtained by selecting a corresponding one of the plurality of candidate ML estimates that maximizes a value of the objective function.

8. The method of claim 6 , wherein a projection matrix is derived from the single training symbol with the L identical portions, the projection matrix having a structure that is exploited to reduce a number of correlations used by the objective function to perform the grid search.

9. The method of claim 1 , wherein the initial estimate and the ML estimate are both obtained without using a radio frequency tone and direct current components as a CFO estimation reference.

10. A system for carrier-frequency-offset (CFO) estimation for coherent optical orthogonal-frequency-division-multiplexing (CO-OFDM) broadband systems, comprising:

an initial carrier frequency offset estimator for obtaining an initial estimate of a normalized CFO with an estimation range equal to ±L/2 subcarrier subspacing using a single training symbol with L identical portions;

a memory for storing the initial estimate; and

a maximum likelihood estimator for obtaining a maximum likelihood (ML) estimate of the CFO by performing a local grid search based on the initial estimate,

wherein the ML estimate is calculated by maximizing over a set of grid search results that fall within a range around the initial estimate defined by an empirically designed parameter, and

wherein the ML estimate is calculated as:

ε

ML

=

arg

⁢

⁢

max



ε

-

ε

^

ini



≤

ε

b

⁢

{

⁢

(

∑

l

=

1

L

-

1

⁢

p

⁡

(

l

)

⁢

ⅇ

-

j

⁢

⁢

2

⁢

π

⁢

⁢

ε

⁢

⁢

l

/

L

)

}

,

where ε is a normalized CFO, {circumflex over (ε)} ini is an initial CFO estimate ε b is an empirically designed parameter, and p(l) provides correlations.

11. The system of claim 10 , wherein the initial estimate is obtained without considering channel information and without a grid search.

12. The system of claim 10 , wherein the initial estimate is obtained by calculating averaged correlation coefficients between time-domain received symbols, calculating a phase difference between any two consecutive ones of the averaged correlation coefficients, calculating a weight factor of the phase difference, and averaging a product of the weight factor and the phase difference.

13. The system of claim 10 , wherein the local grid search is limited to searching only within a pre-specified vicinity of the initial estimate.

14. The system of claim 10 , wherein the local grid search comprises a plurality of grid searches, and a number of the plurality of grid searches performed to obtain the ML estimate is determined by the empirically designed parameter and a grid interval.

15. The system of claim 10 , wherein the ML estimate is obtained by jointly estimating the CFO and an objective function, the objective function for performing the grid search to identify a plurality of candidate ML estimates.

16. The system of claim 15 , wherein the ML estimate is obtained by selecting a corresponding one of the plurality of candidate ML estimates that maximizes a value of the objective function.

17. The system of claim 15 , wherein a projection matrix is derived from the single training symbol with the L identical portions, the projection matrix having a structure that is exploited to reduce a number of correlations used by the objective function to perform the grid search.

18. The system of claim 10 , wherein the initial estimate and the ML estimate are both obtained without using a radio frequency tone and direct current components as a CFO estimation reference.