LDPC Decoder Node Compression for Lower Memory Die Area
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing decoders for low-density parity-check (LDPC) codes require significant memory to store variable and check node states, which increases complexity and die area, but reducing the number of bits per node can degrade performance and limit the number of errors correctable.
Innovation Solution
Implementing a compression module to compress and store variable and check node information in a compressed format, using techniques such as lossy or lossless compression to reduce memory requirements while maintaining decoding performance, and utilizing bit flipping decoders with hybrid message-passing algorithms to efficiently update and decompress data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If memory is increased to store variable and check node states, then decoding performance is improved, but die area and complexity increase
Solution Approach 1:
The patent changes the representation parameters of node states by using exponential encoding where the magnitude of values is represented by the number of trailing zeros and the sign by the most significant bit. This parameter transformation allows the same decoding performance to be achieved with fewer memory bits, directly reducing die area while maintaining reliability.
Solution Approach 2:
The patent applies different storage strategies to different parts of the data structure. Variable nodes and check nodes use different memory allocation and encoding schemes based on their specific requirements. This localized optimization allows memory to be used more efficiently for each node type, reducing overall die area while maintaining the decoding performance required for each node's function.
2Area of stationary object
If number of bits per node is reduced to decrease memory usage, then die area is reduced, but decoding performance and error correction capability are degraded
Solution Approach 1:
The patent transforms the parameter representation from storing actual magnitude values to storing encoded representations where trailing zeros indicate magnitude and the most significant bit indicates sign. This parameter change enables the same error correction capability to be maintained with fewer bits per node, directly addressing the contradiction between memory reduction and performance maintenance.
Solution Approach 2:
The patent creates a compressed representation (copy) of the node state information that preserves the essential decoding functionality. By copying only the critical sign information and deriving magnitude from the encoding structure itself, the system maintains error correction capability while using significantly fewer bits than storing complete magnitude values would require.
3Area of stationary object
If memory is reduced to store compressed node information, then die area and complexity are reduced, but memory access and update operations become more complex
Solution Approach 1:
The patent extracts only the essential information (sign bit and position of trailing zeros) from the complete node state representation. By taking out only the critical components needed for decoding and deriving the rest from the encoding structure, the system reduces memory size while keeping the compression and decompression logic relatively simple and manageable.
Data Source
AI summary
A decoder including a compression module configured to select one or more nodes from a plurality of nodes associated with data being decoded by the decoder, where each node includes one or more bits, and to compress the one or more bits associated with the selected nodes. A memory is configured to store the compressed one or more bits associated with the selected nodes.