System and method for improving convergence during modem training and reducing computational load during steady-state modem operations

• US 5,864,545 A
• Filed: 12/06/1996
• Issued: 01/26/1999
• Est. Priority Date: 12/06/1996
• Status: Expired due to Term

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: AltoCom, Inc. (Broadcom, Inc.)
• Inventors: Gonikberg, Mark, Liang, Haixiang
• Primary Examiner(s): Ton, Dang
• Assistant Examiner(s): Sam, Phirin
• Application Number:
  US08/761,405
• Time in Patent Office: 781 Days
• Field of Search: 370/286, 370/289, 370/290, 370/292, 370/293, 370/296, 370/276, 370/287, 375/231, 375/232, 375/233, 375/234, 375/229, 379/406, 379/410, 379/407, 379/411
• US Class Current: 370/286
• CPC Class Codes: H04B 3/235 combined with adaptive equaliser

Abstract

A phase-splitting T/3 equalizer and echo canceller structure is computationally efficient because only one point per baud is calculated. However, there are two drawbacks to the structure: (1) since the equalizer performs both the phase-splitting function and channel response equalization, its convergence is slow, and (2) when training the echo canceller during half-duplex training, an answering modem needs an assumed equalizer in its receive path to train its echo canceller, because the adaptive equalizer has not yet been trained; however, after equalizer training the echo canceller needs to be retrained because equalizer coefficients have changed. In contrast, a fixed phase splitting filter can be used during training. The echo canceller and equalizer are each trained with the fixed phase splitting filter thereby improving convergence performance, and after training, the equalizer is convolved with the fixed phase splitting filter to provide the combined phase splitting equalizer and the equalizer is convolved with the echo canceller to provide the combined echo canceller. In this way, computational load is small in steady state and convergence is fast during training. Even though the convergence rate is slower in data mode, it suitable for tracking line variations. An advantage of the transformation from the training structure to the steady-state structure is that a modem or system exploiting the technique provides improved convergence during training while reducing computational load during full-duplex operations. Half-duplex applications, i.e., modem or system operations without echo cancellation, similarly benefit from improved convergence during training and reduced computational load during steady-state operations.

Claims

1. A method for training echo canceller and equalizer structures of a modem having transmit and receive paths, the method comprising the steps of:

providing a half duplex training structure including;

an adaptive phase-splitting echo canceller coupled to receive transmit-path samples;

a phase splitting filter coupled to receive receive-path samples; and

an adaptive complex equalizer coupled to the phase-splitting echo canceller and the phase splitting filter to receive a first difference between outputs thereof;

training the phase-splitting echo canceller using the first difference;

decoupling the complex equalizer from the transmit path and thereafter training the complex equalizer using a second difference between an output of the complex equalizer and a reference; and

convolving the trained phase-splitting echo canceller with the trained complex equalizer and convolving the phase splitting filter with the trained complex equalizer to define a trained full-duplex structure, the trained full-duplex structure including;

a trained phase-splitting echo canceller coupled to receive transmit-path samples; and

a trained phase-splitting equalizer coupled to receive receive-path samples.

2. A method as recited in claim 1,

wherein phase-splitting echo canceller training is performed during a transmit-only phase of modem training;

and

wherein complex equalizer training is performed during a receive-only phase of modem training.

3. A method as recited in claim 1, further comprising the step of:

continuing, after the convolving steps, to train the trained phase-splitting echo canceller and the trained phase-splitting equalizer during a full-duplex phase of modem training.

4. A method as recited in claim 1, further comprising the step of:

receiving data using the trained phase-splitting echo canceller and the trained phase-splitting equalizer to compensate for communications channel impairments.

5. A method as recited in claim 1, wherein the trained phase-splitting echo canceller and the trained phase-splitting equalizer defined by the convolving steps are adaptive structures.

6. A method as recited in claim 1, wherein the trained phase-splitting echo canceller and the trained phase-splitting equalizer defined by the convolving steps are fractionally-spaced structures.

7. A method as recited in claim 1, wherein the trained phase-splitting echo canceller and the trained phase-splitting equalizer defined by the convolving steps are adaptive T/3 structures.

8. The method of claim 1 wherein the complex equalizer, the trained phase-splitting echo canceller, and the trained phase-splitting equalizer downsample by an integer amount.

9. The method of claim 1 represented as instructions executable on a general purpose processor and encoded in a machine readable medium.

10. The method of claim 1 represented as digital signal processor (DSP) instructions executable on a DSP.

11. The method of claim 1 wherein at least one of the adaptive phase-splitting echo canceller, the phase splitting filter, the adaptive complex equalizer, the trained phase-splitting echo canceller, and the trained phase-splitting equalizer comprise instructions executable on a general purpose processor and encoded in a machine readable medium.

12. A method for training a modem having a receive path, the method comprising:

providing a half duplex training structure including;

a phase splitting filter coupled to receive receive-path samples; and

an adaptive complex equalizer coupled to receive an output of the phase splitting filter;

training the complex equalizer using a difference between an output of the complex equalizer and a reference; and

convolving the phase splitting filter with the trained complex equalizer to define a trained phase-splitting equalizer coupled to receive receive-path samples.

13. The method of claim 12, wherein the method further includes:

receiving data in steady-state half-duplex mode using the trained phase-splitting equalizer to compensate for communications channel impairments.

14. The method of claim 12,

wherein the modem has a transmit path;

wherein providing further includes providing an adaptive phase-splitting echo canceller coupled to receive transmit-path samples;

wherein the adaptive complex equalizer is further coupled to receive a first difference between an output of the phase-splitting echo canceller and the output of the phase splitting filter; and

wherein the method further includes;

training the phase-splitting echo canceller using the first difference;

decoupling the complex equalizer from the adaptive phase-splitting echo canceller prior to the complex equalizer training; and

convolving the trained phase-splitting echo canceller with the trained complex equalizer to define a trained phase-splitting echo canceller coupled to receive transmit-path samples.

15. The method of claim 14, wherein the method further includes:

receiving data in steady-state full-duplex mode using the trained phase-splitting equalizer to compensate for communications channel impairments.

16. A modem having transmit and receive paths and comprising:

a phase splitting filter coupled into the receive path;

a first phase splitting T/3 echo canceller coupled between the transmit and receive paths during a transmit-only phase of half-duplex training, wherein coefficients of the first phase splitting T/3 echo canceller are updated during the transmit-only phase to minimize a first difference between outputs of the phase splitting filter and the first phase splitting T/3 echo canceller;

a complex T/3 equalizer coupled into the receive path during a receive-only phase of half-duplex training, wherein coefficients of the complex T/3 equalizer are updated during the receive-only phase to minimize a second difference between the output of the complex T/3 equalizer and a reference;

a second phase splitting T/3 echo canceller coupled into the receive path during full-duplex operations, the second phase splitting T/3 echo canceller defined by convolution of respective coefficients of the first phase splitting T/3 echo canceller and the complex T/3 equalizer coincident with a cut-over to full-duplex operations; and

a phase-splitting T/3 equalizer coupled into the receive path during full-duplex operations, the phase-splitting T/3 equalizer defined by convolution of respective coefficients of the phase splitting filter and the complex T/3 equalizer coincident with the cut-over to full-duplex operations.

17. A modem as recited in claim 16,

wherein the complex T/3 equalizer converges faster than the phase splitting T/3 equalizer, thereby improving convergence performance during half-duplex training;

and

wherein the phase splitting T/3 equalizer is computationally more efficient than the complex T/3 equalizer, thereby imposing a lower computational load during full-duplex operations.

18. A modem as recited in claim 16, wherein the fractionally-spaced phase splitting echo canceller is already trained at cut-over to full-duplex operations.

19. The modem of claim 17,

wherein the phase splitting T/3 echo canceller and the phase splitting T/3 equalizer are each represented using instructions executable on a general purpose processor and encoded in a machine readable medium;

and

wherein computational efficiency of the phase splitting T/3 echo canceller and the phase splitting T/3 equalizer allow processing bandwidth of the general purpose processor to be shared with application programs.

20. The modem of claim 16 wherein the phase splitting filter, the first and the second phase splitting T/3 echo cancellers, the phase splitting T/3 equalizer, and the complex T/3 equalizer are each represented as discrete time filters using instructions executable on a processor and encoded in a machine readable medium.

21. The modem of claim 20, wherein the processor includes a general purpose processor, processing bandwidth of which is allocable to the discrete time filters and to application programs.

22. The modem of claim 20, wherein the processor includes a digital signal processor (DSP).

23. An apparatus comprising:

a half-duplex training structure comprising;

adaptive phase-splitting fractionally-spaced echo canceller means;

phase splitting filter means; and

adaptive complex fractionally-spaced equalizer means;

means for training the adaptive phase-splitting fractionally-spaced echo canceller means during a transmit only phase of half-duplex training;

means for training the adaptive complex fractionally-spaced equalizer means during a receive only phase of half-duplex training; and

means for converting the half-duplex training structure to a full-duplex structure, the converting means;

convolving trained coefficients of the adaptive phase-splitting fractionally-spaced echo canceller means with trained coefficients of the adaptive complex fractionally-spaced equalizer means to define trained fractionally-spaced phase-splitting echo canceller means; and

convolving trained coefficients of the phase splitting filter means with trained coefficients of the adaptive complex fractionally-spaced equalizer to define trained phase-splitting fractionally-spaced equalizer means.

24. An apparatus, as recited in claim 23, further comprising:

a processor; and

machine readable storage,

at least one of the adaptive phase-splitting fractionally-spaced echo canceller means, the phase splitting filter means, the adaptive complex fractionally-spaced equalizer means, the adaptive phase-splitting fractionally-spaced echo canceller training means, the adaptive complex fractionally-spaced equalizer training means, the converting means, the trained fractionally-spaced phase-splitting echo canceller means; and

the trained phase-splitting fractionally-spaced equalizer means, comprising instructions stored in the machine readable storage and executable on the processor.

25. An apparatus, as recited in claim 23, encoded in a computer readable medium as instructions executable on a processor.

26. A method for providing fast convergence during training of a modem while limiting steady-state computational load of said modem, said method comprising:

training a precursor training structure having improved convergence characteristics as compared to a phase-splitting equalizer;

after training, transforming from the precursor training structure to a receive path structure including the phase-splitting equalizer structure by convolving filters of the precursor training structure to define the phase-splitting equalizer; and

thereafter receiving samples using the receive path structure.

27. A method, as recited in claim 26, wherein the precursor training structure comprises:

a phase-splitting filter coupled to receive receive-path samples; and

an adaptive complex equalizer coupled to receive an output of the phase-splitting filter.

28. A method, as recited in claim 26, wherein the precursor training structure comprises:

an adaptive phase-splitting echo canceller coupled to receive transmit-path samples;

a phase-splitting filter coupled to receive receive-path samples; and

an adaptive complex equalizer coupled to the phase-splitting echo canceller and the phase-splitting filter to receive a first difference between outputs thereof.

29. A method, as recited in claim 28, wherein the training comprises:

first training the phase-splitting echo canceller using the first difference;

after the first training, decoupling the complex equalizer from the transmit path; and

thereafter second training the complex equalizer using a second difference between an output of the complex equalizer and a reference.

30. A method, as recited in claim 26, wherein the training includes training a complex equalizer of the precursor training structure using a difference between an output of the complex equalizer and a reference.

31. A method, as recited in claim 26,

wherein the precursor training structure is a half-duplex training structure;

and

wherein the receive path structure is a trained full-duplex structure.

32. A method, as recited in claim 26,

wherein the precursor training structure is a half-duplex training structure;

and

wherein the receive path structure is a trained half-duplex structure.

33. A method, as recited in claim 26,

wherein the receive path structure is a full-duplex structure;

and

wherein the filter convolving comprises;

convolving a trained phase-splitting echo canceller of the precursor training structure with a trained complex equalizer of the precursor training structure to define a trained phase-splitting echo canceller of the full-duplex receive path structure; and

convolving a fixed phase-splitting filter of the precursor training structure with the trained complex equalizer to define a trained phase-splitting equalizer of the full-duplex receive path structure.

34. A method, as recited in claim 26,

wherein the receive path structure is a half-duplex structure;

and

wherein the filter convolving comprises convolving a fixed phase-splitting filter of the precursor training structure with a trained complex equalizer of the precursor training structure to define a trained phase-splitting equalizer of the half-duplex receive path structure.