System and method for state reduction in trellis equalizers using bounded state enumeration

Patent metadata

Field	Value
Publication number	US-9660845-B2
Application number	US-201514876378-A
Country	US
Kind code	B2
Filing date	Oct 6, 2015
Priority date	Oct 6, 2015
Publication date	May 23, 2017
Grant date	May 23, 2017

Abstract

Embodiment reduced-state trellis equalization techniques compute accumulated path metrics (APMs) for a subset of candidate states for at least some stages in the trellis based on a neighborhood map of an ML state. This reduces the number of APMs that are computed and stored during trellis equalization. Other embodiments select a subset of candidate states for which APMs are transported to the next stage of the trellis based on the neighborhood map. This eliminates the need to sort the remaining APMs during reduced state trellis equalization. The neighborhood map identifies a subset of the highest probability neighbors for an ML state. The subset of candidate states identified as highest probability neighbors may be saved in a look-up table. The look-up table may be generated offline and/or generated/updated dynamically during run-time operation.

First claim

What is claimed is:

1. A method for reduced state trellis equalization, the method comprising:
receiving, over an interface of a device, a signal carrying at least a current symbol and a trailing symbol over a system exhibiting channel memory, the current symbol being received before the trailing symbol, the current symbol being mapped to a current stage of a trellis and the trailing symbol being mapped to a next stage of the trellis;
selecting, by a controller in the device, a subset of candidate states at the current stage of the trellis, the current stage of the trellis including the subset of candidate states and one or more remaining candidate states that are excluded from the subset of candidate states;
computing, by an equalizer in the device, a first set of accumulated path metrics (APMs) for the subset of candidate states at the current stage of the trellis without computing APMs for the one or more remaining candidate states at the current stage of the trellis such that APMs are computed for fewer than all states at the current stage of the trellis; and
decoding, by the equalizer, the trailing symbol using at least the APMs for the subset of candidate states of the current stage of the trellis.

2. The method of claim 1 , wherein decoding the trailing symbol at the next stage of the trellis using at least the APMs computed for the subset of candidate states of the current stage of the trellis comprises:
propagating the APMs for the subset of candidate states of the current stage of the trellis from the current stage of the trellis to the next stage of the trellis;
computing APMs for at least some candidate states at the next stage of the trellis using the APMs propagated from the current stage of the trellis; and
selecting a best path over the trellis based on the APMs computed at the next stage of the trellis.

3. A method for reduced state trellis equalization, the method comprising:
receiving, over an interface of a device, a signal carrying at least a current symbol and a trailing symbol over a system exhibiting channel memory, the current symbol being received before the trailing symbol, the current symbol being mapped to a current stage of a trellis and the trailing symbol being mapped to a next stage of the trellis;
selecting, by a controller in the device, a subset of candidate states at the current stage of the trellis, the current stage of the trellis including the subset of candidate states and one or more remaining candidate states that are excluded from the subset of candidate states, wherein selecting the subset of candidate states for the current stage of the trellis comprises identifying a maximum likelihood (ML) state for the current stage of the trellis based on measurements of the signal and selecting the subset of candidate states for the current stage of the trellis from a look up table based on the ML state;
computing, by an equalizer in the device, a first set of accumulated path metrics (APMs) for the subset of candidate states at the current stage of the trellis without computing APMs for the one or more remaining candidate states at the current stage of the trellis; and
decoding, by the equalizer, the trailing symbol using at least the APMs for the subset of candidate states of the current stage of the trellis.

4. The method of claim 3 , wherein the look up table identifies highest probability neighbors of the ML state.

5. The method of claim 3 , wherein the ML state is mapped to an effective channel memory hyper-plane, and wherein a neighborhood map of the ML state on the effective channel memory hyper-plane identifies the subset of candidates states as highest probability neighbors of the ML state.

6. The method of claim 5 , wherein a first dimension of the effective channel memory hyper-plane corresponds to an in-phase component value of the current symbol, and a second dimension of the effective channel memory hyper-plane corresponds to an in-phase component value of a leading symbol that precedes the current symbol.

7. The method of claim 6 , wherein the subset of candidate states correspond to combinations of the in-phase component value of the current symbol and the leading symbol that are most likely to produce the ML state.

8. The method of claim 6 , wherein a first dimension of the effective channel memory hyper-plane corresponds to a quadrature component value of the current symbol, and a second dimension of the effective channel memory hyper-plane corresponds to a quadrature component value of a leading symbol that precedes the current symbol.

9. The method of claim 8 , wherein the subset of candidate states correspond to combinations of the quadrature component values of the current symbol and the leading symbol that are most likely to produce the ML state.

10. A device for reduced state trellis equalization, the device comprising:
a receiver configured to receive a signal carrying at least a current symbol and a trailing symbol over a system exhibiting channel memory, the current symbol being received before the trailing symbol, the current symbol being mapped to a current stage of a trellis and the trailing symbol being mapped to a next stage of the trellis;
a controller configured to select a subset of candidate states at the current stage of the trellis, the current stage of the trellis including the subset of candidate states and one or more remaining candidate states that are excluded from the subset of candidate states; and
a trellis equalizer coupled to the receiver and the controller, the trellis equalizer configured to compute a first set of accumulated path metrics (APMs) for the subset of candidate states at the current stage of the trellis without computing APMs for the one or more remaining candidate states at the current stage of the trellis such that APMs are computed for fewer than all states at the current stage of the trellis, and to decode the trailing symbol using at least the APMs for the subset of candidate states of the current stage of the trellis.

11. A method for reduced state trellis equalization, the method comprising:
receiving, by a receiver comprising one or more processors, a current symbol and a trailing symbol over a system exhibiting channel memory, the current symbol being received before the trailing symbol, the current symbol being mapped to a current stage of a trellis and the trailing symbol being mapped to a next stage of the trellis;
computing, by at least one of the processors in the receiver, accumulated path metrics (APMs) for candidate states at the current stage of the trellis;
identifying, by at least one of the processors in the receiver, a maximum likelihood (ML) state of the current stage based on the APMs;
selecting, by at least one of the processors in the receiver, a subset of candidate states based on a neighborhood map of the ML state without sorting APMs of neighboring candidate states of the ML state following identification of the ML state, the current stage of the trellis including the subset of candidate states and one or more remaining candidate states excluded from the subset of candidate states; and
decoding, by at least one of the processors in the receiver, the trailing symbol at the next stage of the trellis using APMs of the subset of candidate states of the current stage of the trellis without using APMs of the one or more remaining candidate states excluded from the subset of candidate states.

12. The method of claim 11 , wherein the subset of candidate states includes the ML state and one or more of the neighboring candidate states identified as highest probability neighbors of the ML state by the neighborhood map.

13. A receiver co

Assignees

• Huawei Tech Co Ltd

Inventors

• Mitra Jeebak
• Zhang Zhuhong

Classifications

• H04L2025/0342

  QAM · CPC title

• H04L25/03203 Primary

  Trellis search techniques · CPC title