Polar Code Decoding with Dynamic LLR Overflow Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The successive cancellation (SC) decoding algorithm for polar codes faces overflow issues due to finite bit-width limitations in Log-Likelihood Ratio (LLR) computations, particularly in the G function operations, which can lead to errors during the decoding process, especially under poor channel quality conditions.
Innovation Solution
The decoding apparatus dynamically adjusts the starting level for overflow processing based on the number of encoded bits, input bit-width, and internal bit-width, and configures LLR values by multiplying them with coefficients to prevent overflow, allowing for dynamic adjustment according to channel quality, thereby postponing or delaying overflow processing when channel fading occurs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If LLR computation is performed using standard SC decoding algorithm, then decoding accuracy is maintained, but overflow occurs due to finite bit-width limitations
Solution Approach 1:
The patent applies preliminary action by determining a starting level for overflow processing before actual decoding occurs. The method calculates the starting level based on the number of encoded bits, input bit-width, and internal bit-width, and prepares overflow handling coefficients in advance. This allows the system to proactively prevent overflow issues before they occur during the decoding process, rather than reacting after overflow has already happened.
Solution Approach 2:
The patent changes parameters by dynamically adjusting the starting level for overflow processing based on channel quality conditions. When channel fading is detected, the system modifies the starting level parameter and applies scaling coefficients to LLR values. This parameter adjustment allows the decoder to adapt to varying channel conditions and prevent overflow while maintaining decoding accuracy under different signal quality scenarios.
2Reliability
If overflow processing is applied at all decoding levels, then LLR overflow is prevented, but computational complexity increases
Solution Approach 1:
The patent applies local quality by implementing overflow processing selectively at specific decoding levels rather than uniformly across all levels. The method determines a starting level based on system parameters and applies overflow handling only from that level onwards. This localized approach ensures that overflow prevention is applied where most needed while avoiding unnecessary computational overhead at levels where overflow is less likely to occur.
Solution Approach 2:
The patent uses partial action by applying overflow processing only to the necessary portion of the decoding process. Instead of processing all LLR values at all levels, the system identifies the starting level where overflow becomes a concern and applies scaling coefficients only from that point forward. This partial application of overflow handling reduces computational complexity while still preventing the harmful effects of overflow.
3Ease of manufacture
If fixed starting level is used for overflow processing, then implementation is simplified, but adaptability to channel quality variations is reduced
Solution Approach 1:
The patent implements dynamics by making the starting level for overflow processing adjustable based on channel quality conditions. Rather than using a fixed starting level, the system can dynamically determine the appropriate starting level based on the number of encoded bits, input bit-width, internal bit-width, and detected channel conditions. This dynamic approach allows the decoder to adapt to varying channel quality while maintaining a systematic method for determining when to apply overflow handling.
Data Source
AI summary
A method and an apparatus for decoding polar codes, the method comprising: determining a starting level for processing an overflow according to a number of encoded bits of a received polar encoded codeword, an input bit-width, and an internal bit-width of a decoder; multiplying an output Log-Likelihood Ratio (LLR) value and two input LLR values of the G function by a first coefficient and a second coefficient respectively; and finally, the LLR values corresponding to the received codeword are decoded to obtain decoded bits.