Parallel-Concatenated LDPC Convolutional Codes for Low-Latency Decoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current LDPC codes, particularly convolutional LDPC codes, face challenges with high complexity, low data rates, and limited flexibility due to their complex design and large block sizes, which increase latency and reduce system flexibility in dynamic wireless communication systems.

Innovation Solution

The development of Parallel-Concatenated Trellis-based Quasi-Cyclic Low Density Parity Check (LDPC) Convolutional Codes, which employ a QC-RSC encoder and TQC-LDPC MAP decoder to achieve power-efficient decoding with fine granularity and reduced complexity, enabling scalable code-length and high throughput.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If large block sizes are used in LDPC codes to approach Shannon capacity, then coding efficiency is improved, but latency increases and system flexibility is reduced

Engineering Contradiction:
Improvecoding efficiencyVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent divides the large block code into smaller sub-blocks that can be processed independently and in parallel. This segmentation allows the system to achieve coding efficiency close to Shannon capacity while reducing latency by processing multiple smaller blocks simultaneously rather than waiting for one large block to complete.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent employs a flexible block structure where the code length can be dynamically adjusted by concatenating multiple sub-blocks. This dynamic approach allows the system to adapt block sizes to specific application requirements, balancing between coding efficiency and latency needs for different communication scenarios.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If convolutional LDPC codes are used to achieve fine granularity, then data rate flexibility is improved, but decoding complexity increases

Engineering Contradiction:
Improvedata rate flexibilityVSAvoiddecoding complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the convolutional decoding process into multiple independent sub-block decoders. Each sub-block can be decoded separately using simplified algorithms, reducing the overall decoding complexity while maintaining the fine granularity and data rate flexibility characteristics of convolutional codes.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent combines multiple simple sub-block decoders to achieve the functionality of a complex convolutional decoder. By merging several low-complexity decoding units that process segmented blocks, the system attains the adaptability of convolutional codes without the full decoding complexity.

Inventive Principle:
Principle #5Merging (Combining)

3Manufacturing precision

If TQC-LDPC convolutional codes are used to achieve fine granularity, then input granularity is improved, but normalized signal to noise ratio performance deteriorates

Engineering Contradiction:
Improveinput granularityVSAvoidnormalized signal to noise ratio
Core Design Contradiction:
Manufacturing precisionVSReliability

Solution Approach 1:

The patent modifies key parameters of the TQC-LDPC code structure, including the parity check matrix construction and lifting factor selection, to simultaneously achieve fine input granularity and improved normalized signal to noise ratio performance. These parameter optimizations allow the code to maintain granular flexibility while correcting the performance degradation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3231094B1Method and apparatus for parallel concatenated LDPC convolutional codes enabling power-efficient decoders
Publication Date: 2023.11.29 SAMSUNG ELECTRONICS CO LTD
  • EP3231094B1 patent drawingFigure 1
  • EP3231094B1 patent drawingFigure 2a
  • EP3231094B1 patent drawingFigure 2b

AI summary

A method of encoding includes receiving input systematic data including an input group (xz(n)) of Z systematic bits. The method includes generating an LDPC base code using the input group (xz(n)). The LDPC base code is characterized by a row weight (Wr), a column weight (Wc), and a first level lifting factor (Z). The method includes transforming the LDPC base code into a Trellis-based Quasi-Cyclic LDPC (TQC-LDPC) convolutional code. The method includes generating a Parallel Concatenated TQC-LDPC convolutional code in a form of an H-matrix including a systematic submatrix (Hsys) of the input systematic data and a parity check submatrix (Hpar) of parity check bits, wherein the Hpar includes a column of Z-group parity bits. The method includes concatenating the Hpar with each column of systematic bits, wherein the Hpar includes J parity bits per systematic bit.