Sparse channel dual-error tracking adaptive filter/equalizer

• US 7,450,518 B2
• Filed: 09/27/2004
• Issued: 11/11/2008
• Est. Priority Date: 03/12/2003
• Status: Expired due to Fees

• Current Assignee: Avago Technologies General IP PTE Limited (Broadcom, Inc.)
• Original Assignee: Broadcom Corporation (Broadcom, Inc.)
• Inventors: Yousef, Nabil R.
• Primary Examiner(s): Ngo; Ricky
• Assistant Examiner(s): Chery; Dady
• Application Number:
  US10/951,318
• Publication Number:
  US 20050053129A1
• Time in Patent Office: 1,506 Days
• Field of Search: 370/216, 370/228, 370/225, 370/242, 370/252, 370/277, 370/278, 370/437, 370/441, 370/442, 370/479, 375/229, 375231-233, 375/346
• US Class Current: 370/242
• CPC Class Codes: H03H 21/0012 Digital adaptive filters
• CPC Class Codes: H04L 2025/03566 between different tapped delay line structures
• CPC Class Codes: H04L 2025/03636 Algorithms using least mean square [LMS]
• CPC Class Codes: H04L 2025/03687 of step size
• CPC Class Codes: H04L 25/03057 with a recursive structure H04L25/03031 takes precedence
• CPC Class Codes: H04L 25/03987 Equalisation for sparse channels

Abstract

An adaptive filter has a plurality of filter coefficients and a filter coefficients processing unit trains of the adaptive filter. The filter coefficients processing unit is operable to iterate to alter the plurality of filter coefficients of the communication system filter. Every Nth iteration, the filter coefficients processing unit is operable to identify a first group of filter coefficients of the plurality of filter coefficients that satisfy filter coefficient threshold rules and a second group of filter coefficients of the plurality of filter coefficients that do not satisfy the filter coefficient threshold rules. Every iteration, the filter coefficients processing unit is operable to: (1) update the first group of filter coefficients using a first step size and a first error signal; and (2) update the second group of filter coefficients using a second step size and a second error signal.

Claims

1. A method for training a communication system filter comprising:

performing an initial estimate of a plurality of filter coefficients of the communication system filter; and

based upon an output produced by the communication system filter in response to a received signal and an expected output of the communication system filter, iterating to alter the plurality of filter coefficients of the communication system filter by;

every Nth iteration, identifying a first group of filter coefficients of the plurality of filter coefficients that satisfy filter coefficient threshold rules and a second group of filter coefficients of the plurality of filter coefficients that do not satisfy the filter coefficient threshold rules by;

comparing the magnitude of each filter coefficient of the plurality of filter coefficients to a filter coefficient threshold value;

when the magnitude of a filter coefficient exceeds the filter coefficient threshold value, assigning the filter coefficient to the first group of filter coefficients; and

when the magnitude of the filter coefficient does not exceed the filter coefficient threshold value, assigning the coefficient to the second group of filter coefficients; and

every iteration;

updating the first group of filter coefficients using a first step size and a first error signal; and

updating the second group of filter coefficients using a second step size and a second error signal.

2. The method of claim 1, further comprising assigning the filter coefficient to the first group of filter coefficients when the magnitude of at least one filter coefficient within a neighbor group of filter coefficients exceeds the threshold value.

3. The method of claim 1, wherein performing an initial estimate of the plurality of filter coefficients of the communication system filter comprises assigning a plurality of default filter coefficients.

4. The method of claim 1, wherein performing an initial estimate of the plurality of filter coefficients of the communication system filter comprises using the received signal to assign a plurality of estimated filter coefficients.

5. The method of claim 1, farther comprising, every Nth iteration:

determining the first step size so that it is inversely proportional to a number of filter coefficients in the first group of filter coefficients; and

determining the second step size.

6. The method of claim 5, wherein the second step size is inversely proportional to one of:

a number of filter coefficients in the second group of filter coefficients; and

a number of the plurality of filter coefficients.

7. The method of claim 1 wherein:

updating the first group of filter coefficients using the first step size and the first error signal includes;

determining a first filter output in response to the received signal by using only the first group of filter coefficients; and

determining the first error signal based upon the first filter output and the expected output; and

updating the second group of filter coefficients using the second step size and the second error signal;

determining a second filter output in response to the received signal using all of the plurality of filter coefficients; and

determining the second error signal based upon the second filter output and the expected output.

8. The method of claim 1, wherein the communication system filter is one of a linear equalizer, a portion of a decision feedback equalizer, and a decision feedback equalizer.

9. The method of claim 1, wherein the iterating is performed using one of a Least Mean Squares (LMS) algorithm and a modified LMS algorithm.

14. The method of claim 1, further comprising, every Nth iteration:

determining the first step size so that it is inversely proportional to a number of filter coefficients in the first group of filter coefficients; and

determining the second step size.

15. The method of claim 14, wherein the second step size is inversely proportional to one of:

a number of filter coefficients in the second group of filter coefficients; and

a number of the plurality of filter coefficients.

10. A method for training a communication system filter comprising:

performing an initial estimate of a plurality of filter coefficients of the communication system filter; and

iterating to alter the plurality of filter coefficients of the communication system filter by;

every Nth iteration;

identifying a first group of filter coefficients of the plurality of filter coefficients that satisfy filter coefficient threshold rules and a second group of filter coefficients that do not satisfy the filter coefficient threshold rules by;

comparing the magnitude of each filter coefficient of the plurality of filter coefficients to a filter coefficient threshold value;

when the magnitude of a filter coefficient exceeds the filter coefficient threshold value, assigning the filter coefficient to the first group of filter coefficients; and

when the magnitude of the filter coefficient does not exceed the filter coefficient threshold value, assigning the coefficient to the second group of filter coefficients;

determining a first step size that it is inversely proportional to a number of filter coefficients in the first group of filter coefficients; and

determining a second step size; and

every iteration;

determining a first filter output in response to the received signal using only the first group of filter coefficients;

determining a first error signal based upon the first filter output and an expected output;

updating the first group of filter coefficients using upon the first step size and the first error signal;

determining a second filter output in response to the received signal using all of the plurality of filter coefficients;

determining a second error signal based upon the second filter output and the expected output; and

updating the second group of filter coefficients using the second step size and the second error signal.

11. The method of claim 10, further comprising assigning the filter coefficient to the first group of filter coefficients when the magnitude of at least one filter coefficient within a neighbor group of filter coefficients exceeds the threshold value.

12. The method of claim 10, wherein performing an initial estimate of the plurality of filter coefficients of the communication system filter comprises assigning a plurality of default filter coefficients.

13. The method of claim 10, wherein performing an initial estimate of the plurality of filter coefficients of the communication system filter comprises using the received signal to assign a plurality of estimated filter coefficients.

16. The method of claim 10, wherein the communication system filter is one of a linear equalizer, a portion of a decision feedback equalizer, and a decision feedback equalizer.

17. The method of claim 10, wherein the iterating is performed using one of a Least Mean Squares (LMS) algorithm and a modified LMS algorithm.

18. A communication system filter comprising:

an adaptive filter having a plurality of filter coefficients that produces an output in response to a received signal; and

a filter coefficients processing unit operable to cause adaptive filter training of the adaptive filter, the filter coefficients processing unit operable to iterate to alter the plurality of filter coefficients of the communication system filter by;

every Nth iteration, identifying a first group of filter coefficients of the plurality of filter coefficients that satisfy filter coefficient threshold rules and a second group of filter coefficients of the plurality of filter coefficients that do not satisfy the filter coefficient threshold rules by;

comparing the magnitude of each filter coefficient of the plurality of filter coefficients to a filter coefficient threshold value;

when the magnitude of a filter coefficient exceeds the filter coefficient threshold value, assigning the filter coefficient to the first group of filter coefficients; and

when the magnitude of the filter coefficient does not exceed the filter coefficient threshold value, assigning the coefficient to the second group of filter coefficients; and

every iteration;

updating the first group of filter coefficients using a first step size and a first error signal; and

updating the second group of filter coefficients using a second step size and a second error signal.

19. The communication system filter of claim 18, wherein the filter coefficients processing unit is further operable to assign the filter coefficient to the first group of filter coefficients when the magnitude of at least one filter coefficient within a neighbor group of filter coefficients exceeds the threshold value.

20. The communication system filter of claim 18, wherein every Nth iteration, the filter coefficients processing unit is operable to:

determine the first step size so that it is inversely proportional to a number of filter coefficients in the first group of filter coefficients; and

determine the second step size.

21. The communication system filter of claim 18, wherein every Nth iteration, the filter coefficients processing unit is operable to:

determine the first step size so that it is inversely proportional to a number of filter coefficients in the first group of filter coefficients; and

determine the second step size so that it is inversely proportional to a number of filter coefficients in the second group of filter coefficients.

22. The communication system filter of claim 18 wherein every iteration, the filter coefficients processing unit is operable to:

update the first group of filter coefficients using the first step size and the first error signal by;

determining a first filter output in response to the received signal by using only the first group of filter coefficients; and

determining the first error signal based upon the first filter output and the expected output; and

update the second group of filter coefficients using the second step size and the second error signal by;

determining a second filter output in response to the received signal using all of the plurality of filter coefficients; and

determining the second error signal based upon the second filter output and the expected output.

23. The communication system filter of claim 18, wherein the communication system filter is one of a linear equalizer, a portion of a decision feedback equalizer, and a decision feedback equalizer.

24. The communication system filter of claim 18, wherein the filter coefficients processing unit is operable to iterate using one of a Least Mean Squares (LMS) algorithm and a modified LMS algorithm.