Method and device for calculating coefficients of feed-forward equalizer and feed-backward equalizer in decision feedback equalizer

1. A method for calculating a feed-forward equalizer (FFE) coefficient of an FFE in a minimum mean square error decision feedback equalizer (MMSE-DFE) based on a fast transversal recursive squares (FT-RLS) algorithm, the FFE having a length L F , L F being a positive integer, the method comprising:

a) generating a channel impulse response (CIR) estimation vector h according to an input signal of the FFE;

b) generating a priori forward prediction error φ according to the CIR estimation vector h and a forward prediction coefficient vector w;

c) providing a minimum cost of forward prediction α;

d) generating a normalized gain vector c according to the forward prediction coefficient vector w, the minimum cost of forward prediction α and the priori forward prediction error φ, wherein the normalized gain vector c includes elements c[0] to c[L F ];

e) providing a posteriori forward prediction error ψ

f) updating the forward prediction coefficient vector w according to the posteriori forward prediction error ψ and the normalized gain vector c;

g) providing a conversion factor γ;

h) generating the FFE coefficient according to the conversion factor γ and the normalized gain vector c; and

updating the FFE with said FFE coefficient;

wherein, step (a) to step (g) are collectively a 1 st iteration operation and have an L F number of iterations; step (d) is a 2 nd iteration operation having an n number of iterations, and wherein in the N th iteration of the L F iterations of the 1 st iteration operation, the number of iterations of the 2 nd iteration operation n equals N−1; the 2 nd iteration operation computes the elements c[1] to c[N−1], and directly sets the elements c[N] to c[L F ] to 0 instead of computing the elements c[N] to c[L F ].

2. The method according to claim 1 , wherein in the 1 st iteration of the L F iterations of the 1 st iteration operation, the elements c[1] to c[L F ] are not computed and are directly set to 0.

3. The method according to claim 1 , wherein step (f) is a 3 rd iteration operation having an m number of iterations, where m is an integer between 0 and (L F −1).

4. The method according to claim 3 , wherein in the M th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 3 rd iteration operation m is an integer between (M−1) and L F .

5. The method according to claim 4 , wherein the forward prediction coefficient vector w includes elements w[0] to w[L F −1]; in the M th iteration of the L F iterations of the 1 st iteration operation, the 3 rd iteration operation comprises computing the elements w[0] to w[M−2], where M is an integer greater than 1.

6. The method according to claim 5 , wherein in the M th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 3 rd iteration operation m equals M−1; the 3 rd iteration operation computes the elements w[0] to w[M−2], and directly sets the elements w[M−1] to w[L F −1] to 0 instead of computing the elements w[M−1] to w[L F −1].

7. The method according to claim 6 , in the 1 st iteration of the L F iterations of the 1 st iteration operation, the elements w[0] to w[L F ] are not computed and are directly set to 0.

8. The method according to claim 1 , wherein the CIR estimation vector h includes an L F number of elements and has a length L CIR , the length L CIR is a positive integer and is smaller than L F ; step (b) is a 4 th iteration operation having a p number of iterations; in the P th iteration of the L F iterations of the 1 st iteration operation, and the iteration number of the 4 th iteration operation p is an integer between L CIR to (P−2), where P is an integer between (L CIR +2) and L F .

9. The method according to claim 8 , wherein the CIR estimation vector h includes elements h[0] to h[L F −1]; the elements h[0] to h[ CIR −1] are greater than a predetermined value, and the elements h[L CIR ] to h[L F −1] are smaller than the predetermined value; in the P th iteration of the L F iterations of the 1 st iteration operation, the 4 th iteration operation comprises computing according to the elements h[0] to h[L CIR −1], where P is an integer between (L CIR +2) and L F .

10. The method according to claim 9 , wherein in the P th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 4 th iteration operation p equals L CIR , where P is an integer between (L CIR +2) and L F ; the 4 th iteration operation computes according to only the elements h[0] to h[L CIR −1] and not according to any of the elements h[L CIR −1] to h[L F −1].

11. A feed-forward equalizer (FFE) coefficient calculating device, for calculating an FFE coefficient of an FFE in a minimum mean square error decision feedback equalizer (MMSE-DFE) based on a fast transversal recursive squares (FT-RLS) algorithm, the FFE having a length L F , L F being a positive integer, the FFE coefficient calculating device comprising:

a priori forward prediction error calculating module, generating a priori forward prediction error φ according to a channel impulse response (CIR) estimation vector h and a forward prediction coefficient vector w;

a minimum forward prediction cost calculating module, providing a minimum cost of forward prediction α;

a normalized gain vector calculating module, generating a normalized gain vector c according to the forward prediction coefficient vector w, the minimum cost of forward prediction α, and the priori forward prediction error φ, wherein the normalized gain vector c includes elements c[0] to c[L F ];

a posteriori forward prediction error calculating module, providing a posteriori forward prediction error ψ;

a forward prediction coefficient vector calculating module, updating the forward prediction coefficient vector w according to the posteriori forward prediction error ψ and the normalized gain vector c;

a conversion factor calculating module, providing a conversion factor γ; and

a control module, controlling an operation sequence of the above modules to complete a 1 st iteration operation, generating the FFE coefficient according to the conversion factor γ and the normalized gain vector c, and updating the FFE with said FFE coefficient;

wherein, the 1 st iteration operation has an L F number of iterations, and the normalized gain vector calculating module performs a 2 nd iteration operation having an n number of iterations, and wherein in the N th iteration of the L F iterations of the 1 st iteration operation, the number of iterations of the 2 nd iteration operation n equals N−1; the 2 nd iteration operation computes the elements all to c[1] to c[N−1], and directly sets the elements c[N] to c[L F ] to 0 instead of computing the elements c[N] to c[L F ].

12. The FFE coefficient calculating device according to claim 11 , wherein in the 1 st iteration of the L F iterations of the 1 st iteration operation, the normalized gain vector calculating module directly sets the elements c[1] to c[L F ] to 0 instead of computing the elements c[1] to c[L F ].

13. The FFE coefficient calculating device according to claim 11 , wherein the forward prediction coefficient vector calculating module performs a 3 rd iteration operation having an m number of iterations, where m is an integer between 0 and (L F −1).

14. The FFE coefficient calculating device according to claim 13 , wherein in the M th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 3 rd iteration operation m is an integer between (M−1) and L F .

15. The FFE coefficient calculating device according to claim 14 , wherein the forward prediction coefficient vector w includes elements w[0] to w[L F −1]; in the M th iteration of the L F iterations of the 1 st iteration operation, the forward prediction coefficient vector calculating module computes the elements w[0] to w[M−2] in the 3 rd iteration operation, where M is an integer greater than 1.

16. The FFE coefficient calculating device according to claim 15 , wherein in the M th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 3 rd iteration operation m equals M−1; the forward prediction coefficient vector calculating module computes the elements w[0] to w[M−2] in the 3 rd iteration operation, and directly sets the elements w[M−1] to w[L F −1] to 0 instead of computing the elements w[M−1] to w[L F −1].

17. The FFE coefficient calculating device according to claim 16 , wherein in the 1 st iteration of the L F iterations of the 1 st iteration operation, the forward prediction coefficient vector calculating module directly sets the elements w[0] to w[L F ] to 0 instead of computing the elements w[0] to w[L F ].

18. The FFE coefficient calculating device according to claim 11 , wherein the CIR estimation vector h includes an L F number of elements and has a length L CIR , the length L CIR is a positive integer and is smaller than L F ; the priori forward prediction error calculating module performs a 4 th iteration operation having a p number of iterations; in the P th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 4 th iteration operation p is an integer between L CIR and (P−2), where P is an integer between (L CIR +2) and L F .

19. The FFE coefficient calculating device according to claim 18 , wherein the CIR estimation vector h includes elements h[0] to h[L F −1]; the elements h[0] to h[L CIR −1] are greater than a predetermined value, and the elements h[L CIR ] to h[L F −1] are smaller than the predetermined value; in the P th iteration of the L F iterations of the 1 st iteration operation, the priori forward prediction error calculating module computes according to the elements h[0] to h[L CIR −1] in the 4 th iteration operation, where P is an integer between (L CIR +2) and L F .

20. The FFE coefficient calculating device according to claim 19 , wherein in the P th iteration of the L F iterations of the 1 st iteration operation, the iteration number of the 4 th iteration operation p equals L CIR , where P is an integer between (L CIR +2) and L F ; the priori forward prediction error calculating module computes according to only the elements h[0] to h[L CIR −1] and not according to any of the elements h[L CIR −1] to h[L F −1].