System, device, and method for time-domain equalizer training using a two-pass auto-regressive moving average model

1. A channel shortening method for training a time domain equalizer comprising:

determining a first shortened channel impulse response for a communication channel using a first channel modeling scheme;

forming a time-mirrored image of the first shortened channel impulse response;

determining a second shortened channel impulse response for the time-mirrored image of the first shortened channel impulse response using a second channel modeling scheme;

combining the first shortened channel impulse response with an inverse of the second shortened channel impulse response to obtain a third shortened channel impulse response; and

employing the third shortened channel impulse response to configure a time domain equalizer.

2. The channel shortening method of claim 1 , wherein determining a first shortened channel impulse response for a communication channel using a first channel modeling scheme comprises applying an auto-regressive moving average model using p 1 poles and q 1 zeros to form the first shortened channel impulse response having a first approximation H 1 (z)=B 1 (z)/1+A 1 (z), wherein q 1 is greater than a predetermined cyclic prefix length.

3. The channel shortening method of claim 1 , wherein determining a second shortened channel impulse response for me time-mirrored image of the first shortened channel impulse response using a second channel modeling scheme comprises applying an auto-regressive moving average model using p 2 poles and q 2 zeros to form the second shortened channel impulse response having a second approximation H 1 (z)=B 1 (z)/1+A 1 (z), wherein q 1 is less than or equal to a predetermined cyclic prefix length.

4. The channel shortening method of claim 1 , wherein the first shortened channel impulse response has a first approximation H 1 (z)=B 1 (z)/1+A 1 (z) and the second shortened channel impulse response has a second approximation H 2 (z)=B 2 (z)/1+A 2 (z), and wherein combining the first shortened channel impulse response with an inverse of the second shortened channel impulse response to obtain a third shortened channel impulse response comprises combining A 1 (z) and A 2 (1/z) with appropriate delay.

5. The channel shortening method of claim 1 , wherein the first channel modeling scheme is based upon a pole-zero model having p 1 poles and the second channel modeling scheme is based upon a pole-zero model having p 2 poles and wherein the number of taps in a channel shortening filter equals (p 1 +p 2 +1).

6. The method of claim 1 , wherein the communication channel is an Asymmetric Digital Subscriber Line (ADSL) upstream channel.

7. An apparatus comprising a time-domain equalizer for equalizing a communication channel and training logic for training the time-domain equalizer based upon a training signal received over the communication channel, wherein the training logic is operably coupled to determine a set of coefficients for the time-domain equalizer using a two-pass auto-regressive moving average model.

8. The apparatus of claim 7 , wherein the training logic comprises:

first channel modeling logic operably coupled to determine a first shortened channel impulse response for the communication channel;

inversion logic operably coupled to form a time-mirrored image of the first shortened channel impulse response;

second channel modeling logic operably coupled to determine a second shortened channel impulse response for the time-mirrored image of the first shortened channel impulse response; and

coefficient determination logic operably coupled to combine the first shortened channel impulse response with an inverse of the second shortened channel impulse response to obtain a third shortened channel impulse response.

9. The apparatus of claim 8 , wherein the first channel modeling logic is operably coupled to apply an auto-regressive moving average model using p 1 poles and q 1 zeros to form the first shortened channel impulse response having a first approximation H 1 (z)=B 1 (z)/1+A 1 (z), wherein q 1 is greater than a predetermined cyclic prefix length.

10. The apparatus of claim 8 , wherein the second channel modeling logic is operably coupled to apply an auto-regressive moving average model using p 2 poles and q 2 zeros to form the second shortened channel impulse response having a second approximation H 2 (z)=B 2 (z)/1+A 2 (z), wherein q 2 is less than or equal to a predetermined cyclic prefix length.

11. The apparatus of claim 8 , wherein the first shortened channel impulse response has a first approximation H 1 (z)=B 1 (z)/1+A 1 (z) and the second shortened channel impulse response has a second approximation H 2 (z)=B 2 (z)/1+A 2 (z), and wherein the coefficient determination logic is operably coupled to combine A 1 (z) and A 2 (1/z)) with appropriate delay in order to obtain the third shortened channel impulse response.

12. The apparatus of claim 8 , wherein the first channel modeling logic is based upon a pole-zero model having p 1 poles and the second channel modeling logic is based upon a pole-zero model having p 1 poles, and wherein the number of taps the time-domain equalizer equals (p 1 +p 1 +1).

13. The apparatus of claim 7 , wherein the communication channel is an Asymmetric Digital Subscriber Line (ADSL) upstream channel, and wherein the apparatus is a central ADSL terminal unit.

14. A program product including a computer readable medium having stored therein a computer program, the computer program for training a time-domain equalizer based upon a training signal received over a communication channel, the program product comprising:

first channel modeling logic programmed to determine a first shortened channel impulse response for the communication channel;

inversion logic programmed to form a time-mirrored image of the first shortened channel impulse response;

second channel modeling logic programmed to determine a second shortened channel impulse response for the time-mirrored image of the first shortened channel impulse response;

coefficient determination logic programmed to combine the first shortened channel impulse response with an inverse of the second shortened channel impulse response to obtain a third shortened channel impulse response; and

configuration logic operable to employ the third shortened channel impulse response to configure a time domain equalizer.

15. The program product of claim 14 , wherein the first channel modeling logic is programmed to apply an auto-regressive moving average model using p 1 poles and q 1 zeros to form the first shortened channel impulse response having a first approximation H 1 (z)=B 1 (z)/1+A 1 (z), wherein q 1 is greater than a predetermined cyclic prefix length.

16. The program product of claim 14 , wherein the second channel modeling logic is programmed to apply an auto-regressive moving avenge model using p 2 poles and q 2 zeros to form the second shortened channel impulse response having a second approximation H 2 (z)=B 2 (z)/1+A 2 (z), wherein q 2 is less than or equal to a predetermined cyclic prefix length.

17. The program product of claim 14 , wherein the first shortened channel impulse response has a first approximation H 1 (z)=B 1 (z)/1+A 1 (z) and the second shortened channel impulse response has a second approximation H 2 (z)=B 2 (z)/1+A 2 (z), and wherein the coefficient determination logic is programmed to combine A 1 (z) and A 2 (1/z) with appropriate delay in order to obtain the third shortened channel impulse response.

18. The program product of claim 14 , wherein the first channel modeling logic is based upon a pole-zero model having p 1 poles and the second channel modeling logic is based upon a pole-zero model having p 1 poles, and wherein the number of taps the time-domain equalizer equals (p 1 +p 2 +1).

19. The program product of claim 14 , wherein the communication channel is an Asymmetric Digital Subscriber Line (ADSL) upstream channel.