LDPC Base Matrix Layout for Shared Layered Decoding Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Low Density Parity Check (LDPC) codes used in 5G wireless communications face challenges in efficiently managing large block sizes and high throughput requirements, particularly in achieving low energy consumption and reducing the complexity of switching networks in LDPC decoders.
Innovation Solution
The LDPC base matrix is adapted to comprise non-identical, row-orthogonal parts, with column-wise combinations derived from a common starting vector through cyclic shifting or interleaving, allowing for efficient encoding and decoding while reducing the complexity of switching networks and energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional LDPC base matrices are used with separate switching networks for each layer, then decoding accuracy is maintained, but device complexity and chip area increase due to multiple switching networks
Solution Approach 1:
The patent merges multiple layer-specific switching networks into a single shared switching network by designing the base matrix with row-orthogonal parts that have column-wise combinations representing cyclically shifted versions of the same starting vector. This allows the same switching network to be reused across multiple layers with different cyclic shift values, reducing device complexity while maintaining decoding accuracy through the mathematical properties of the base matrix structure.
Solution Approach 2:
The switching network is designed to be universal and multi-functional, serving multiple layers simultaneously. The base matrix structure enables a single switching network to handle different layers by applying different cyclic shift values to the same starting vector, making the switching network versatile and eliminating the need for separate dedicated networks for each layer.
2Use of energy by stationary object
If multiple identical switching networks are implemented for different layers, then layer-specific decoding is achieved, but energy consumption increases due to redundant operations
Solution Approach 1:
The patent combines multiple layer-specific switching networks into one shared network, eliminating redundant operations across layers. The base matrix structure with row-orthogonal parts allows the same switching network to serve multiple layers through cyclic shifts, reducing energy consumption while preserving layer-specific decoding capabilities through mathematical transformations rather than physical duplications.
Solution Approach 2:
The patent uses parameter changes (cyclic shift values) to differentiate layer-specific decoding operations instead of requiring separate physical networks. By varying the cyclic shift parameter applied to the same starting vector, the system achieves layer-specific decoding with a single switching network, significantly reducing energy consumption associated with redundant hardware operations.
3Quantity of substance
If complete base matrices are stored for each configuration, then decoding accuracy is ensured, but data storage efficiency decreases
Solution Approach 1:
The patent extracts only the essential starting vector from the complete base matrix structure, storing merely this compact representation. The full base matrix can be regenerated on-demand through cyclic shifts and column-wise combinations of the stored vector, dramatically reducing storage memory usage while ensuring decoding accuracy is maintained through the mathematical reconstruction process.
Solution Approach 2:
The patent performs preliminary action by pre-storing only the compact starting vector that contains all necessary information to generate any required base matrix configuration. This preliminary storage of the essential element enables on-demand regeneration of complete matrices through cyclic shifts, reducing storage requirements while ensuring accurate decoding through systematic reconstruction.
Data Source
AI summary
A base matrix is applied to an LDPC coder. The base matrix includes multiple parts, each including multiple of rows and columns, and containing integers, each representative of an identity matrix cyclically shifted in accordance with the integer or representative of an all-zero matrix. At least two of the multiple parts are configured such that their respective column-wise combinations of rows represents a same starting vector, cyclically shifted or interleaved, with zero or more but not all integers not indicative of the all-zero matrix of the same vector substituted by integers indicative of the all-zero matrix. The at least two of the multiple parts are not identical. The applied base matrix is used for one of encoding data using the LDPC coder or decoding data using the LDPC coder.