Decision feedback equalizer employing fixed ratio postcursor taps for minimizing noise and intersymbol interference in signals conveyed over high speed data service loop

• US 5,414,733 A
• Filed: 12/20/1993
• Issued: 05/09/1995
• Est. Priority Date: 12/20/1993
• Status: Expired due to Term

• Current Assignee: ADTRAN Incorporated (ADTRAN Holdings, Inc.)
• Original Assignee: ADTRAN Incorporated (ADTRAN Holdings, Inc.)
• Inventors: Turner, Michael D.
• Primary Examiner(s): Chin, Stephen
• Assistant Examiner(s): TSE, YOUNG TOI
• Application Number:
  US08/170,412
• Time in Patent Office: 505 Days
• Field of Search: 375/12-14, 375/94, 375/101, 375/103, 333/18, 333/28 R, 364/724.16, 364/724.19, 364/724.2
• US Class Current: 375/233
• CPC Class Codes: H04L 25/03057 with a recursive structure H04L25/03031 takes precedence

Abstract

The feedforward filter section of a decision feedback equalizer is modified to include one or more postcursor taps, that are sequentially weighted at decreasing binary fractions of the cursor tap. Such a modified feedforward filter section, combined with the placement of a simple anti-aliasing filter upstream of the sampling point, results in an optimum feedforward filter configuration that is not anticausal, and offers a substantially improved performance over conventional DFB equalizer structures. Optimum performance is achieved when such a postcursor filter structure is augmented with an adaptive noise canceler coupled in the DFB path.

Claims

1. A decision feedback equalizer comprising:

a linear transversal filter section to which successive received symbols of an input symbol sequence are sequentially introduced, said linear transversal filter section including a feedforward multistage delay line comprised of a series of 1st-Mth symbol value storage stages, M being an integer greater than 1, and a feedforward weighting and summing stage operative to multiply respective ones of a plurality of 1st-Mth weighting coefficients W₁ -W_M by respective ones of a plurality of 1st-Mth received symbols stored in said feedforward multistage delay line, and thereby obtain a plurality of 1st-Mth products, said 1st-Mth products being summed to provide a weighted and summed feedforward symbol value output;

a symbol decision unit which is coupled to receive a differential symbol value associated with a difference between said weighted and summed feedforward symbol value output and a weighted and summed feedback symbol value output and being operative to generate a received symbol value estimate in accordance with said difference; and

a decision feedback transversal filter section to which successive symbol value estimates generated by said symbol decision unit are sequentially applied, said decision feedback transversal filter section including a feedback multistage delay line comprised of a series of 1st through Nth symbol value storage stages, N being an integer greater than 1, and a feedback weighting and summing stage operative to multiply respective ones of a plurality of 1st through Nth weighting coefficients by respective ones of a plurality of 1st through Nth symbol decision outputs stored in said feedback multistage delay line, and thereby obtain a plurality of 1st through Nth products, said 1st through Nth products being summed to provide a weighted and summed feedback output, said weighted and summed feedback output being differentially combined with said weighted and summed feedforward symbol value output of said linear transversal filter section and the resulting differential signal being applied to said symbol decision unit which outputs received symbol value estimates; and

wherein

said feedforward weighting and summing stage of said linear transversal filter section has its largest valued, cursor weighting coefficient W_C associated with an M-Kth symbol stage of said feedforward multistage delay line, where K is an integer equal to or greater than 1, and K respective postcursor weighting coefficients W_M-K to W_M associated with the last K symbol stages of said feedforward multistage delay line having fractional weighting coefficients of said cursor weighting coefficients W_C.

2. The decision feedback equalizer according to claim 1, wherein said respective postcursor weighting coefficients W_M-K to W_M associated with the postcursor weighting coefficient values associated with the last K symbol stages of said feedforward multistage delay line are fixed fractions of said largest valued, cursor weighting coefficient W_C associated with said M-Kth received symbol stage of said feedforward multistage delay line.

3. The decision feedback equalizer according to claim 2, wherein the postcursor weighting coefficient values associated with said last K stages of said feedforward multistage delay line are progressively decreasing binary fractions (1/2)W_C, (1/2)² W_C, . . . , (1/2)^K W_C of said largest valued, cursor weighting coefficient W_C associated with said M-Kth received symbol of said feedforward multistage delay line.

4. The decision feedback equalizer according to claim 1, wherein K equals four.

5. The decision feedback equalizer according to claim 1, further including an adaptive noise canceler coupled to receive said weighted and summed feedforward symbol value output and said received symbol value estimates, and being operative to modify said differential signal applied to said symbol decision unit in accordance with the difference between said weighted and summed feedforward symbol value output and said received symbol value estimates.

6. The decision feedback equalizer according to claim 5, wherein said adaptive noise canceler is operative to subtract said received symbol value estimates from said weighted and summed feedforward symbol value output to provide a noise error value, said noise error value being applied to a noise prediction operator, the output of which is subtracted from said differential signal to be applied to said symbol decision unit.

7. The decision feedback equalizer according to claim 1, further including a precursor estimate unit, coupled to said feedforward weighting and summing stage of said linear transversal filter section and said decision feedback transversal filter section, and being operative to form an estimate of a precursor symbol, and wherein said estimate of the precursor symbol is summed with the output of said feedback weighting and summing stage of said feedback transversal filter prior to being differentially combined with the feedforward weighted and summed symbol value output of said linear transversal filter section for application to said symbol decision unit.

8. The decision feedback equalizer according to claim 7, wherein said precursor estimate unit is operative to provisionally equalize and scale the output of said feedforward weighting and summing stage of said linear transversal filter section by the weighting coefficient associated with said precursor symbol.

9. The decision feedback equalizer according to claim 8, wherein said input symbol sequence is comprised of a plurality of respectively different symbols, and wherein said precursor estimate unit is operative to provisionally equalize and scale the output of said feedforward weighting and summing stage of said linear transversal filter section by respective weighting coefficients associated with each of said respectively different symbols, and wherein said decision feedback transversal filter section is operative to generate a received symbol value estimates as that symbol associated with the precursor estimate having minimum error.

10. The decision feedback equalizer according to claim 8, wherein said input symbol sequence is comprised of a plurality of respectively different symbols, and wherein said precursor estimate unit is operative to provisionally equalize and scale the output of said feedforward weighting and summing stage of said linear transversal filter section by respective weighting coefficients associated with each of said respectively different symbols, and further including a precursor symbol decision unit which is coupled to receive the equalized and scaled outputs of the weighting and summing stage of said linear transversal filter section and to generate a precursor estimate associated with one of said respectively different symbols, and wherein said precursor estimate is combined with said weighted and summed feedback output.

11. The decision feedback equalizer according to claim 1, wherein said linear transversal filter section is coupled to receive fractionally spaced samples of successive ones of said input symbol sequence.

12. A method of processing successive received symbols of a symbol sequence received over a dispersive communication channel comprising the steps of:

(a) sampling said received symbol sequence to produce a sequence of symbol samples;

(b) coupling said sequence of symbol samples to a linear transversal filter section, said linear transversal filter section including a feedforward multistage delay line comprised of a series of 1st-Mth symbol value storage stages, M being an integer greater than 1, and a feedforward weighting and summing stage operative to multiply respective ones of a plurality of 1st-Mth weighting coefficients W₁ -W_M by respective ones of a plurality of 1st-Mth received symbol samples stored in said feedforward multistage delay line, and thereby obtain a plurality of 1st-Mth products, said 1st-Mth products being summed together to provide a feedforward weighted and summed symbol value output;

(c) providing a symbol decision unit which is operative to receive a differential symbol value associated with a difference between said feedforward weighted and summed symbol value output and a weighted and summed feedback symbol value output, and generating a received symbol value estimate in accordance with said difference;

(d) coupling respective received symbol value estimates generated by said symbol decision unit in step (c) to a decision feedback transversal filter section, said decision feedback transversal filter section including a feedback multistage delay line comprised of a series of 1st through Nth symbol value storage stages, and a feedback weighting and summing stage operative to multiply respective ones of a plurality of 1st through Nth weighting coefficients by respective ones of a plurality of 1st through Nth symbol decision outputs stored in said feedback multistage delay line, N being an integer greater than 1, and thereby obtain a plurality of 1st through Nth products, said 1st through Nth products being summed to provide a feedback weighted and summed output;

(e) differentially combining the feedback weighted and summed output provided in step (d) with said feedforward weighted and summed symbol value output provided by said linear transversal filter section in step (b) and applying the resulting differential signal being to said symbol decision unit provided in step (b) so that said symbol decision unit outputs received symbol value estimates; and

wherein

step (b) comprises establishing the weighting coefficients of said feedforward weighting and summing stage of said linear transversal filter section such that the largest valued, cursor weighting coefficient W_C is associated with an M-Kth symbol stage of said feedforward multistage delay line, where K is an integer equal to or greater than 1, and K respective postcursor weighting coefficients W_M-K to W_M associated with the last K symbol stages of said feedforward multistage delay line having fractional weighting coefficients of said cursor weighting coefficients W_C.

13. The method according to claim 12, wherein said respective postcursor weighting coefficients W_M-K to W_M associated with the postcursor weighting coefficient values associated with the last K symbol stages of said feedforward multistage delay line are fixed fractions of said largest valued, cursor weighting coefficient W_C associated with said M-Kth received symbol stage of said feedforward multistage delay line.

14. The method according to claim 13, wherein the postcursor weighting coefficient values associated with said last K stages of said feedforward multistage delay line are progressively decreasing binary fractions (1/2)W_C, (1/2)² W_C, . . . , (1/2)^K W_C of said largest valued, cursor weighting coefficient W_C associated with said M-Kth received symbol of said feedforward multistage delay line.

15. The method according to claim 12, wherein K equals four.

16. The method according to claim 12, further including the step of (f) coupling said feedforward weighted and summed symbol value output and said received symbol value estimates to an adaptive noise canceler which is operative to modify said differential signal applied to said symbol decision unit in accordance with the difference between said feedforward weighted and summed symbol value output and said received symbol value estimates.

17. The method according to claim 16, wherein, in step (f), said adaptive noise canceler is operative to subtract said received symbol value estimates from said weighted and summed feedforward symbol value output to provide a noise error value, and further including applying said noise error value to a noise prediction operator, the output of which is subtracted from said differential signal to be applied to said symbol decision unit.

18. The method according to claim 12, further including the step of (f) deriving an estimate of a precursor symbol in accordance with said feedforward weighted and summed symbol value output and said weighted and summed feedback output, and summing said estimate of said precursor symbol with the feedback weighted and summed output of said decision feedback transversal filter section prior to being differentially combined with the feedforward weighted and summed symbol value output of said linear transversal filter section for application to said symbol decision unit.

19. The method according to claim 18, wherein step (f) comprises provisionally equalizing and scaling the output of said feedforward weighting and summing stage of said linear transversal filter section by the weighting coefficient associated with said precursor symbol.

20. The method according to claim 19, wherein said input symbol sequence is comprised of a plurality of respectively different symbols, and wherein step (f) comprises provisionally equalizing and scaling the output of said feedforward weighting and summing stage of said linear transversal filter section by respective weighting coefficients associated with each of said respectively different symbols, and wherein said decision feedback transversal filter section is operative to generate a received symbol value estimates as that symbol associated with the precursor estimate having minimum error.

21. The method according to claim 19, wherein said input symbol sequence is comprised of a plurality of respectively different symbols, and wherein step (f) comprises provisionally equalizing and scaling the output of said feedforward weighting and summing stage of said linear transversal filter section by respective weighting coefficients associated with each of said respectively different symbols, and further including applying the equalized and scaled outputs of the feedforward weighting and summing stage of said linear transversal filter section to a precursor symbol decision unit which is operative to generate a precursor estimate associated with one of said respectively different symbols, and combining said precursor estimate with said weighted and summed feedback output.

22. The method according to claim 12, wherein step (a) comprises sampling said received symbol sequence to produce a sequence of fractionally spaced samples of successive ones of said input symbols, and step (b) comprises coupling said sequence of fractionally spaced samples of successive ones of said input symbols to said linear transversal filter section.