Methods and apparatus for map detection with reduced complexity
View Patent ↗Methods and apparatus are provided for high-speed, low-power, high-performance channel detection. A soft-output detector is provided for processing a received signal, comprising: a forward detector for calculating forward state metrics; a backward detector for calculating backward state metrics; and a current branch detector for calculating a current branch metric, wherein at least two of the forward detector, the backward detector and the current branch detector employ trellis structures with a different number of states. A method is provided for processing a received signal using a soft-output detector, comprising: calculating forward state metrics using a forward detector; calculating backward state metrics using a backward detector; and calculating a current branch metric using a current branch detector, wherein at least two of the forward detector, the backward detector and the current branch detector employ trellis structures with a different number of states.
1. A soft-output detector for processing a received signal, comprising:
a forward detector for calculating forward state metrics;
a backward detector for calculating backward state metrics; and
a current branch detector for calculating a current branch metric, wherein at least two of said forward detector, said backward detector and said current branch detector employ trellis structures with a different maximum number of states.
2. The soft-output detector of claim 1 , wherein said forward detector uses a four state, radix 2 N trellis with or without feedback, with N≧2, where N is a number of received samples.
3. The soft-output detector of claim 1 , wherein said backward state metrics are unraveled by setting initial backward state metrics to 0 and calculating a best path leaving each current backward state in a forward direction.
4. The soft-output detector of claim 1 , wherein forward branch metrics are used by said forward detector to calculate said forward state metrics and also used by said backward detector to calculate at least a portion of said backward state metrics, and wherein the soft-output detector further comprises one or more registers for delaying said forward state metrics.
5. The soft-output detector of claim 1 , wherein each backward state metric represents a best path of length Q leaving a current backward state in a forward direction, where Q is one of N and 2N, and wherein the forward detector uses a radix 2 N trellis with N≧2, where N is a number of received samples.
6. The soft-output detector of claim 1 , wherein log-likelihood ratios are calculated for a symbol using a combined trellis structure that contains all possible values of the symbol, wherein a combined metric for each possible symbol value is calculated by adding a forward state metric, a current branch metric, and a backward state metric and wherein a dependence on parts of the backward state label, current branch label, and forward state label is removed that do not directly map to the symbol value, wherein a hard-decision is the symbol value associated with the best combined metric in a set of combined metrics for all possible symbol values, wherein a magnitude of the log-likelihood ratio for each possible symbol value is obtained by substantially taking a difference between a best combined metric associated with that particular symbol value and a best combined metric in said set of combined metrics for all possible values.
7. The soft-output detector of claim 1 , wherein one or more of a max-log-MAP algorithm, a log-MAP algorithm, a feedback algorithm, and an optimization method is substantially applied piecewise within said forward, backward and current branch detectors and said soft-output detector to remove a dependence on parts of the received signal that are not constrained by backward state labels, forward state labels, and current branch metric labels.
8. The soft-output detector of claim 1 , wherein a soft-output for one of binary and non-binary symbols are mapped from sets of bits constrained by the labels of the combined trellis structure.
9. The soft-output detector of claim 8 , wherein the bits are precoded with 1/(1⊕D) and wherein the log-likelihood ratios are calculated for an unprecoded binary symbol that corresponds to a (1⊕D) mapping of two precoded bits.
10. The soft-output detector of claim 1 , wherein said backward detector has a reduced number of states at some trellis step.
11. A method for processing a received signal using a soft-output detector, comprising:
calculating forward state metrics using a forward detector;
calculating backward state metrics using a backward detector; and
calculating a current branch metric using a current branch detector, wherein at least two of said forward detector, said backward detector and said current branch detector employ trellis structures with a different maximum number of states.
12. The method of claim 11 , wherein the forward detector uses a four state, radix 2 N trellis with or without feedback, where N is a number of received samples.
13. The method of claim 11 , further comprising the step of unraveling said backward state metrics by setting initial backward state metrics to 0 and calculating a best path leaving each current backward state in a forward direction.
14. The method of claim 11 , wherein each backward state metric represents a best path of length Q leaving a current backward state in a forward direction, where Q is one of N and 2N, and wherein the forward detector uses a radix 2 N trellis with N≧2, where N is a number of received samples.
15. The method of claim 11 , wherein said log-likelihood ratios are calculated for a symbol using a combined trellis structure that contains all possible values of the symbol, wherein a combined metric for each possible symbol value is calculated by adding a forward state metric, a current branch metric, and a backward state metric and wherein a dependence on parts of the backward state label, current branch label, and forward state label is removed that do not directly map to the symbol value, wherein a hard-decision is the symbol value associated with the best combined metric in a set of combined metrics for all possible symbol values, wherein a magnitude of the log-likelihood ratio for each possible symbol value is obtained by substantially taking a difference between a best combined metric associated with that particular symbol value and a best combined metric in said set of combined metrics for all possible values.
16. The method of claim 11 , wherein one or more of a max-log-MAP algorithm, a log-MAP algorithm, a feedback algorithm, and an optimization method is substantially applied piecewise within said forward, backward and current branch detectors and said soft-output detector to remove a dependence on parts of the received signal that are not constrained by backward state labels, forward state labels, and current branch metric labels.
17. The method of claim 11 , wherein a soft-output for one of binary and non-binary symbols are mapped from sets of bits that are constrained by labels of the combined trellis structure.
18. The method of claim 17 , wherein the bits are precoded with 1/(1⊕D) and wherein the log-likelihood ratios are calculated for an unprecoded binary symbol that corresponds to a (1⊕D) mapping of two precoded bits.
19. The method of claim 11 , wherein said backward detector has a reduced number of states at some trellis step.
20. The method of claim 11 , further comprising the step of generating one or more log-likelihood ratios for one or more symbols based on said forward state metrics, backward state metrics, and current branch metric.