LDPC Iterative Decoder Using Sign-Preserving Min-Sum Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current decoders for LDPC codes face challenges in achieving good performance with low computational complexity, particularly in terms of quantization bits, leading to suboptimal decoding efficiency in communication standards.
Innovation Solution
The Sign-Preserving Min-Sum (SP-MS) decoder uses a sign-preserving factor to ensure that messages never propagate erased values, employing a sign-and-magnitude representation to maintain message reliability and improve decoding convergence, even with fewer quantization bits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If Min-Sum or Offset Min-Sum decoders are used to reduce computational complexity, then decoder complexity is reduced, but decoding performance degrades in the waterfall region
Solution Approach 1:
The patent changes the parameter representation from standard LLR values to sign-preserving factors with sign-and-magnitude representation. This transformation modifies how messages are encoded and transmitted between nodes, allowing the simplified Min-Sum algorithm to maintain better performance by preserving sign information that would otherwise be lost in low-precision quantization.
Solution Approach 2:
The patent combines multiple representation schemes (sign-preserving factors, sign-and-magnitude representation, and quantization levels) into a composite message format. This composite approach integrates the advantages of different representation methods to achieve both low complexity and good performance, effectively creating a hybrid solution that outperforms individual approaches.
2Device complexity
If bit-size representation of message is reduced to further reduce decoder complexity, then computational complexity is reduced, but decoding performance is lost
Solution Approach 1:
The patent transitions from representing messages in a single dimension (standard LLR values) to a two-dimensional sign-and-magnitude representation. This dimensional change allows the system to preserve both sign and magnitude information separately, enabling more efficient use of limited bit-width while maintaining message precision through the structured representation format.
Solution Approach 2:
The patent fundamentally changes the parameter representation from continuous LLR values to discrete sign-preserving factors with specific quantization levels. This parameter transformation enables the system to operate effectively with reduced bit-size by mapping continuous values to a structured discrete set that preserves essential information.
3Reliability
If 6 bits of quantization is used to achieve optimal performance, then decoding performance is optimized, but computational complexity increases
Solution Approach 1:
The patent changes the quantization parameter structure from uniform 6-bit representation to a variable sign-and-magnitude format with fewer bits. This parameter reorganization allows the system to achieve comparable performance with reduced bit-width by allocating bits more efficiently between sign and magnitude components, thereby reducing overall complexity.
Solution Approach 2:
The patent employs a simplified message representation that uses fewer bits (cheaper) compared to standard 6-bit quantization. While individual messages use less precision, the sign-preserving structure ensures that critical information is maintained, allowing the system to achieve good performance with lower-cost (fewer bits) message representations.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An iterative decoder configured for decoding a code having a codeword length N, comprises: • N variable nodes (VNs) vn, n = 1...N, configured to receive a LLR In defined on a alphabet A I of qch quantization bits, qch ≥2; • M constraint nodes (CNs) cm, m = 1...M, 2 ≤ M < N; • vn and cm exchanging messages along edges of a Tanner graph; • each vn sending messages m vn → cm toc m , the set of connected constraint nodes being noted V(vn), and V(vn)\{cm} being V(vn) except cm, and, • each cm sending messages m cm → vn to vn; • the LLR In and the messages m vn → cm and m cm→vn are coded according to a sign-and-magnitude code; and • each variable node vn, for each iteration l, compute: • sign-preserving factors: (formula I) where ξ is a positive or a null integer; •(formula II) and (formula III) where S is a function from the set of value that can take floor (formula IV) to the set A s .