Method and apparatus for determining scaling factors in fixed-point decoder
View Patent ↗Data from a communications channel is decoded by receiving data bits corresponding to encoded data, determining a set of data representations from the data bits, distributing the set of data representations into bins across a dynamic range to generate a distribution of the data representations, assigning a respective intermediate scale factor to each bin, deriving a set of moments from the intermediate scale factors, combining the moments into a scaling factor, scaling the data representations by the scaling factor, and sending the scaled data representations to a decoder. The data representations may be a histogram or cumulative distribution function of log-likelihood ratios (LLRs) or values based on channel estimates. In an iterative implementation performed until a stopping condition is met, the data representations may be scaled down on later iterations to avoid saturation. A correction factor may be applied to update the scaling factor for later data bits.
1. A method of decoding data representations derived from a communications channel, the method comprising:
receiving data bits corresponding to encoded data in the communications channel;
determining a set of data representations from the data bits;
distributing the set of data representations into bins across a dynamic range to generate a distribution of the data representations;
assigning a respective intermediate scale factor to each bin;
deriving a set of moments from the intermediate scale factors;
combining the moments into a scaling factor;
scaling the data representations by the scaling factor; and
sending the scaled data representations to a decoder.
2. The method of claim 1 wherein the determining a set of data representations from the data bits comprises determining log-likelihood ratio (LLR) values from the data bits.
3. The method of claim 1 wherein the determining a set of data representations from the data bits comprises determining values from the data bits based on channel estimates.
4. The method of claim 1 wherein the distributing the set of data representations comprises generating a histogram.
5. The method of claim 1 wherein the distributing the set of data representations comprises generating a cumulative distribution function.
6. The method of claim 1 further comprising, when the data bits are encoded with a low-density parity-check (LDPC) code, adjusting the distribution based on parameters of the data bits.
7. The method of claim 6 wherein the adjusting the distribution based on parameters of the data bits comprises adjusting the distribution based on at least one of (a) a number of shortening bits per codeword, (b) a number of punctured bits per codeword, and (c) a number of repeated bits per codeword.
8. The method of claim 1 wherein the combining the moments comprises combining the moments in a weighted summation.
9. The method of claim 1 wherein the combining the moments comprises applying the moments to a machine learning model.
10. The method of claim 1 further comprising:
monitoring channel characteristics of different channels; wherein:
at least one of the assigning, the deriving and the combining is performed differently for different channels according to the channel characteristics; and
at least one channel has a different scaling factor than other channels.
11. The method of claim 1 further comprising:
discarding data representations below a noise threshold; wherein:
the intermediate scale factors are based on noise variance.
12. The method of claim 1 wherein the combining the moments into a scaling factor comprises:
grouping the bins into ranges; and
deriving separate scaling factors for each range.
13. The method of claim 1 wherein the method is iterative and comprises scaling down the data representations on later iterations to avoid saturation.
14. The method of claim 1 further comprising applying a correction factor to update the scaling factor for later data bits in a packet.
15. Apparatus for decoding data representations derived from a communications channel, the apparatus comprising:
a receiver configured to receive data bits corresponding to encoded data in the communications channel;
processing circuitry configured to:
determine a set of data representations from the data bits,
distribute the set of data representations into bins across a dynamic range to generate a distribution of the data representations,
assign a respective intermediate scale factor to each bin,
derive a set of moments from the intermediate scale factors,
combine the moments into a scaling factor, and
scale the data representations by the scaling factor; and
decoding circuitry configured to decode scaled data representations.
16. The apparatus of claim 15 wherein the processing circuitry is configured to determine a set of log-likelihood ratio (LLR) values from the data bits.
17. The apparatus of claim 15 wherein the processing circuitry is configured to determine a set of values from the data bits based on channel estimates.
18. The apparatus of claim 15 wherein the processing circuitry is configured to, when the data bits are encoded with a low-density parity-check (LDPC) code, adjust the distribution based on parameters of the data bits.
19. The apparatus of claim 18 wherein the processing circuitry is configured to adjust the distribution based on at least one of (a) a number of shortening bits per codeword, (b) a number of punctured bits per codeword, and (c) a number of repeated bits per codeword.
20. The apparatus of claim 18 wherein the processing circuitry is configured to combine the moments in a weighted summation.
21. The apparatus of claim 18 wherein the processing circuitry is configured to apply the moments to a machine learning model.
22. The apparatus of claim 15 wherein the processing circuitry is further configured to:
monitor channel characteristics of different channels; wherein:
the processing circuitry is configured to perform at least one of the assigning, the deriving and the combining is performed differently for different channels according to the channel characteristics; and
the processing circuitry is configured to assign a different scaling factor at least one channel as compared to other channels.
23. The apparatus of claim 15 wherein the processing circuitry is further configured to:
discard data representations below a noise threshold; and
base the intermediate scale factors on noise variance.
24. The apparatus of claim 15 wherein the processing circuitry is further configured to combine the moments into a scaling factor by:
grouping the bins into ranges; and
deriving separate scaling factors for each range.
25. The apparatus of claim 15 wherein the processing circuitry is further configured to:
operate iteratively; and
scale down the data representations on later iterations to avoid saturation.
26. The apparatus of claim 15 wherein the processing circuitry is further configured to apply a correction factor to update the scaling factor for later data bits in a packet.
27. A method of decoding data representations derived from a communications channel, the method comprising:
receiving data bits corresponding to encoded data in the communications channel;
determining a set of data representations from the data bits;
scaling the data representations by a scaling factor to fit a dynamic range;
sending the scaled data representations to a decoder;
iteratively performing the sending until a stop condition is met; and
scaling down the data representations on later iterations to avoid saturation.