Sliding window turbo decoder
First Claim
1. A method of reducing calculations in the decoding of a received convolutionally coded signal represented by a trellis of a predetermined block length, the method comprising the steps of:
- a) dividing the trellis into windows;
b) selecting a first window of the trellis having a known first state metric;
c) decoding a portion of the trellis using backward recursion starting from a point that is after the end of the window selected in the previous step backwards to the end of the window, defining a learning period, to determine a known state metric at the end of the window, wherein a length of the learning period is dependent on the quality of the signal such that a shorter learning period is chosen for a higher quality signal and a longer learning period is chosen for a low er quality signal;
d) decoding a portion of the trellis within the window using forward and backward recursion starting from the respective known state metrics at a beginning and end of the window defined in the previous step so as to determine the forward and backward recursion state metrics at each stage in the window;
e) calculating a soft output at each stage of the window using the forward recursion state metrics, the branch metrics, and the stored backward recursion state metrics;
f) determining the quality of the signal from the previous step;
g) adjusting the learning period to be shorter as the quality of the signal improves and longer if the quality of the signal worsens; and
h) selecting a next window of the trellis and proceeding with the steps c)-g) until the entire trellis is decoded.
4 Assignments
0 Petitions
Accused Products
Abstract
Decoding signals represented by a trellis of a block length divided into windows includes a step of decoding a portion of the trellis using backward recursion starting from a point that is after the end of a window backwards to the end of the window, defining a learning period, to determine a known state metric at the end of the window. A length of the learning period for each window dependents on the signal quality such that a shorter learning period is chosen for a higher signal quality. The signal quality used is an intrinsic signal-to-noise ratio derived from the log-likelihood-ratio of the soft outputs of the decoded window. In particular, the intrinsic signal-to-noise ratio of the signal is defined as a summation of generated extrinsic information multiplied by a log-likelihood-ratio (LLR) value at each iteration.
-
Citations
20 Claims
-
1. A method of reducing calculations in the decoding of a received convolutionally coded signal represented by a trellis of a predetermined block length, the method comprising the steps of:
-
a) dividing the trellis into windows;
b) selecting a first window of the trellis having a known first state metric;
c) decoding a portion of the trellis using backward recursion starting from a point that is after the end of the window selected in the previous step backwards to the end of the window, defining a learning period, to determine a known state metric at the end of the window, wherein a length of the learning period is dependent on the quality of the signal such that a shorter learning period is chosen for a higher quality signal and a longer learning period is chosen for a low er quality signal;
d) decoding a portion of the trellis within the window using forward and backward recursion starting from the respective known state metrics at a beginning and end of the window defined in the previous step so as to determine the forward and backward recursion state metrics at each stage in the window;
e) calculating a soft output at each stage of the window using the forward recursion state metrics, the branch metrics, and the stored backward recursion state metrics;
f) determining the quality of the signal from the previous step;
g) adjusting the learning period to be shorter as the quality of the signal improves and longer if the quality of the signal worsens; and
h) selecting a next window of the trellis and proceeding with the steps c)-g) until the entire trellis is decoded. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 19, 20)
-
-
13. A method of reducing calculations in the decoding of a received convolutionally coded sequence of signals represented by a trellis of a predetermined block length, the method comprising the steps of:
-
a) dividing the trellis into windows;
b) selecting a first window of the trellis having a known first state metric;
c) decoding a portion of the trellis using backward recursion starting from a point that is after the end of the window selected in the previous step backwards to the end of the window, defining a learning period, to determine a known state metric at the end of the window, wherein a length of the learning period is dependent on an intrinsic signal-to-noise ratio of the signal such that the learning period is shortened as the intrinsic signal-to-noise ratio of the signal improves;
d) decoding a portion of the trellis within the window using forward and backward recursion starting from the respective known state metrics at a beginning and end of the window defined in the previous step so as to determine the forward and backward recursion state metrics at each stage in the window;
e) calculating a soft output at each stage of the window using the forward recursion state metrics, the branch metrics, and the stored backward recursion state metrics;
f) determining the intrinsic signal-to-noise ratio of the signal defined as a summation of generated extrinsic information multiplied by a log-likelihood-ratio (LLR) value at each iteration generated in the second decoding step;
g) adjusting the learning period from an initial upper boundary to be shorter as the quality of the signal improves but not more than a lower boundary; and
h) selecting a next window of the trellis and proceeding with the steps c)-g) until the entire trellis is decoded.
-
-
14. A radiotelephone with a receiver and demodulator with a soft-decision output decoder for serially processing windows of a convolutionally coded signal, represented by a trellis of predetermined block length divided into windows in a frame buffer, the soft-decision output decoder comprising:
-
a memory;
a learning recursion processor decodes a portion of the trellis using a learning backward recursion from a point that is after the end of a window backward to the end of the window, defining a learning period, to determine a known state metric at the end of the window;
a backward recursion processor subsequently decodes the portion of the trellis within the window using backward recursion from the known state at the end of the window back to the beginning of the window to define a set of known backward recursion state metrics within the window which can be stored in the memory;
a forward recursion processor decodes the portion of the trellis within the window using forward recursion starting from a known state at the beginning of the window and moving forward to define a set of known forward recursion state metrics within the window which can be stored in the memory; and
a decoder coupled to the memory calculates a soft output at each stage of the window using the forward and backward recursion state metrics and branch metrics at each stage, the decoder also determines a quality of the signal for each window and adjusts the learning period for processing a next window to be shorter as the quality of the signal improves and longer if the quality of the signal worsens. - View Dependent Claims (15, 16, 17, 18)
-
Specification