Blind training of a decision feedback equalizer

• US 6,069,917 A
• Filed: 05/23/1997
• Issued: 05/30/2000
• Est. Priority Date: 05/23/1997
• Status: Expired due to Term

• Current Assignee: Avago Technologies General IP PTE Limited (Broadcom, Inc.)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Yang, Jian, Werner, Jean-Jacques
• Primary Examiner(s): Chin, Stephen
• Assistant Examiner(s): Fan, Chieh M.
• Application Number:
  US08/862,909
• Time in Patent Office: 1,103 Days
• Field of Search: 375/229, 375/232, 375/233, 375/321, 375/230, 375/234, 364/724.19, 607/17, 708/322, 708/323
• US Class Current: 375/233
• CPC Class Codes: H04L 25/0307 using blind adaptation

Abstract

A decision feedback equalizer (DFE) comprises a feed-forward filter and a feedback filter. Blind training of the DFE is performed using a statistical-based tap updating algorithm for the feed-forward filter, and a symbol-based type of tap updating algorithm for the feedback filter.

Claims

1. A method for blindly converging an equalizer having a feed-forward portion and a feedback portion, the method comprising the steps of:

using a statistical-based equalization technique for converging the feed-forward portion and

using a symbol-based equalization technique for converging the feedback portion;

wherein the statistical-based equalization technique is based on a multimodulus algorithm (MMA).

2. The method of claim 1 wherein the symbol-based equalization technique is based on a least-mean square (LMS) algorithm.

3. The method of claim 1 wherein the symbol-based equalization technique is based on a fourth-order least-mean square (LMS) algorithm.

4. A method for converging an equalizer having a feed-forward portion and a feedback portion, the method comprising the steps of:

generating a signal as a function of output signals of the feed-forward portion and the feedback portion;

symbol slicing the generated signal for generating a first sequence of symbols taken from a constellation comprising N symbols;

symbol slicing the generated signal for generating a second sequence of symbols taken from a constellation having M symbols, where N<

M;

using values of the first sequence of symbols for converging the feed-forward portion; and

using values of the second sequence of symbols for converging the feedback portion.

5. The method of claim 4 wherein the step of using values for converging the feed-forward portion includes the step of operating on these values in accordance with a reduced constellation-based algorithm (RCA).

6. The method of claim 4 wherein the step of using values for converging the feedback portion includes the step of operating on these values in accordance with a least-mean square (LMS)-based algorithm.

7. The method of claim 4 wherein the step of using values for converging the feedback portion includes the step of operating on these values in accordance with a fourth-order least-mean square (LMS)-based algorithm.

8. A method for blindly converging a decision feedback equalizer having a feed-forward portion and a feedback portion, the method comprising the steps of:

generating a signal as a function of output signals of the feed-forward portion and the feedback portion;

symbol slicing the generated signal for generating a sequence of symbols taken from a constellation having M symbols;

using the generated signal as a feedback signal for converging the feed-forward portion; and

using values of the sequence of symbols for converging the feedback portion;

wherein the step of using the generated signal for converging the feed-forward portion includes the step of using a multimodulus-based algorithm (MMA).

9. The method of claim 8 wherein the step of using values for converging the feedback portion includes the step of operating on these values in accordance with a least-mean square (LMS)-based algorithm.

10. The method of claim 8 wherein the step of using values for converging the feedback portion includes the step of operating on these values in accordance with a fourth-order least-mean square (LMS)-based algorithm.

11. Apparatus comprising:

memory for storing program data and tap coefficients for use in a feed-forward filter and a feedback filter; and

a processor for executing the stored program for blindly converging the values of the tap coefficients for a) the feed-forward filter by using a statistical-based blind equalization technique, b) the feedback filter by using a symbol-based equalization technique;

wherein the statistical-based equalization technique is based on a multimodulus algorithm (MMA).

12. The apparatus of claim 11 wherein the symbol-based equalization technique is based on a least-mean square (LMS) algorithm.

13. The apparatus of claim 11 wherein the symbol-based equalization technique is based on a fourth-order least-mean square (LMS) algorithm.

14. Apparatus comprising:

a feedback filter;

a feed-forward filter;

an N-symbol slicer operative on an applied signal, which is developed as a function of output signals from the feedback filter and the feed-forward filter, for generating a first sequence of symbols taken from a constellation comprising N symbols;

an M-symbol slicer operative on the applied signal for generating a second sequence of symbols taken from a constellation having M symbols, where N<

M;

wherein the feed-forward filter adapts as a function of values of the first sequence of symbols and the feedback filter adapts as a function of values of the second sequence of symbols.

15. The apparatus of claim 14 wherein the feed-forward filter adapts as a function of a blind equalization algorithm and the feedback filter adapts as a function of a symbol-based algorithm.

16. The apparatus of claim 15 wherein the symbol-based algorithm is a least-mean square (LMS)-based algorithm.

17. The apparatus of claim 15 wherein the symbol-based algorithm is a fourth-order least-mean square (LMS)-based algorithm.

18. The apparatus of claim 14 wherein the feed-forward filter and the feedback filter form a decision feedback equalizer.

19. The apparatus of claim 14 wherein the feed-forward filter and the feedback filter form a noise predictive decision feedback equalizer.