Granted Patent B2

US 7,471,904 · App. 10/627,098 · Granted Dec 30, 2008

Abstract

Claims (38)

1. A receiver for receiving an optical signal carrying a sequence of data thereon, comprising:

a photo-detector connected to an optical path, carrying said optical signal, for converting said optical signal to an electrical signal having non-Gaussian noise therein; and

an equalizer for removing intersymbol interference and said non-Gaussian noise from said electrical signal, said equalizer having a plurality of coefficients configured to be updated based upon a first least-mean 2N th -order (LMN) algorithm, where N is greater than one.

2. The receiver of claim 1 , further comprising a coefficient controller to update said coefficients based upon said first least-mean 2N th -order (LMN) algorithm, where N is greater than one.

3. The receiver of claim 2 , wherein said equalizer comprises a finite impulse response filter configured to produce a first output signal responsive to said electrical signal, said first output signal being representative of a sum of the associated electrical signal plus a weighted sum of previous ones of the electrical signal, wherein the previous signals are weighted by said coefficients.

4. The receiver of claim 3 , further comprising:

a slicer to produce a predicted signal for each first output signal received from the finite impulse response filter;

a subtractor to produce an error signal proportional to the difference between said first output signal and a corresponding predicted signal or training signal,

wherein said coefficient controller is configured to update said coefficients responsive to the error signal.

5. The receiver of claim 4 , wherein said slicer is configured to produce the predicted signal by adaptively determining a slicing threshold.

6. The receiver of claim 4 , wherein said equalizer is a feed forward equalizer and said coefficient controller is configured to update a set of said coefficients {right arrow over (c)}(k+1) at a time (k+1) as {right arrow over (c)}(k)+βN[e(k)] 2N−1 {right arrow over (u)}(k), wherein β is a preset step size, {right arrow over (c)}(k) and e(k) are respective sets of coefficients and error signals at a time k, and {right arrow over (u)}(k) is an input signal at the time k.

7. The receiver of claim 3 , further comprising:

a first subtractor to produce a second output signal, said second output signal being a sum of one of the first output signals and a corresponding feedback signal;

a slicer to produce a predicted signal in response to each second output signal;

a second subtractor to produce an error signal representing a difference between the second output signal and a corresponding training signal or predicted signal;

a feedback filter to produce the feedback signal in response to corresponding ones of the predicted or training signals, the feedback signal being a weighted sum of the predicted or training signal, wherein weights in the sum being characteristics of the filter; and

a weight controller to update the weights in response to the error signal, the weight controller configured to perform the updates based upon a second least-mean 2N th -order (LMN) algorithm where N is greater than one.

8. The receiver of claim 7 , wherein said equalizer is a decision feedback equalizer and said weight controller is configured to update a set of the weights {right arrow over (w)}(k+1) at a time (k+1) as {right arrow over (w)}(k)+βN[e(k)] 2N−1 {right arrow over (r)}(k), wherein β is a preset step size, {right arrow over (w)}(k) and e(k) are respective sets of weight and error signals at a time k, and {right arrow over (r)} T (k)=[{right arrow over (u)}(k),−{right arrow over (a)}(k)], where {right arrow over (u)}(k) is an input signal at the time k, and {right arrow over (a)}(k) is a predicted or training signal at the time k.

9. The receiver of claim 2 , wherein the equalizer is a digital filter.

10. The receiver of claim 1 , wherein the equalizer is an analog filter.

11. The receiver of claim 1 , wherein said non-Gaussian noise is substantially described by a first component linearly proportional to a noise distribution in said optical signal and a second component proportional to the square of said noise distribution.

12. A method for receiving an optical signal, comprising:

converting said optical signal to an electrical signal having non-Gaussian noise therein;

removing intersymbol interference and said non-Gaussian noise from said electrical signal using an equalizer, wherein said equalizer is configured by a plurality of coefficients; and

updating said plurality of coefficients based upon a least-mean 2N th -order (LMN) algorithm where N is greater than one.

13. The method of claim 12 , wherein said equalizer comprises a finite impulse response filter that is further configured to produce a first output signal responsive to said electrical signal, said first output signal being representative of a sum of the associated electrical signal plus a weighted sum of previous ones of the electrical signal, wherein the previous signals are weighted by said coefficients.

14. The method of claim 13 , further comprising the steps of: producing a predicted signal for each first output signal received from the finite impulse response filter;

producing an error signal proportional to the difference between said first output signal and a corresponding one of the predicted signals or a corresponding training signal; and

updating said coefficients responsive to the error signal.

15. The method of claim 14 , further comprising the step of updating a set of the coefficients {right arrow over (c)}(k+1) at a time (k+1) as {right arrow over (c)}(k)+βN[e(k)] 2N−1 {right arrow over (u)}(k), wherein β is a preset step size, {right arrow over (c)}(k) and e(k) are respective set of coefficients and error signals at a time k, and {right arrow over (u)}(k) is an input signal at the time k.

16. The method of claim 13 , further comprising the steps of:

producing a second output signal, said second output signal being a sum of one of the first output signals and a corresponding feedback signal;

producing a predicted signal in response to each second output signal;

for a particular one of said second output signals, producing an error signal representing a difference between a particular one of said second output signals and a corresponding training signal or predicted signal;

producing the feedback signal in response to corresponding ones of the predicted or training signals, the feedback signal being a weighted sum of the predicted or training signal, wherein weights in the sum being characteristics of the filter; and

updating the weights in response to the error signal with a weight controller configured to perform the updates based upon a Least-mean 2 N th -order (LMN) algorithm where N is greater than one.

17. The method of claim 16 , further comprising the step of updating a set of the weights {right arrow over (w)}(k+1) at a time (k+1) as {right arrow over (w)}(k)+βN[e(k)] 2N−1 {right arrow over (r)}(k), wherein β is a preset step size, {right arrow over (w)}(k) and e(k) are respective sets of weight and error signals at a time k, and {right arrow over (r)} T (k)=[{right arrow over (u)}(k),−{right arrow over (a)}(k)], where {right arrow over (u)}(k) is an input signal at the time k, and {right arrow over (a)}(k) is a predicted or training signal at the time k.

18. The method of claim 14 , wherein said non-Gaussian noise is substantially described by a first component linearly proportional to a noise distribution in said optical signal and a second component proportional to the square of said noise distribution.

Assignments (12)