Bit-Scale Memory Correction via Redundant Byte Replacement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory correcting methods, such as ECC and Redundancy methods, require excessive extra correcting bits, leading to high consumption and inefficient use of resources, as they often replace entire bytes due to a single defective bit, resulting in reduced memory density and increased fabrication costs.
Innovation Solution
A bit-scale memory correcting method that adds a correcting byte with fewer correcting bits than conventional methods, allowing these bits to replace underperforming bits within memory bytes, thereby reducing the overall number of correcting bits needed and enhancing memory density, while also enabling the use of these bits as parity bits for ECC to improve reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional ECC or Redundancy methods are used to correct defective bits, then reliability is improved, but the number of correcting bits increases significantly
Solution Approach 1:
The patent divides the correction resource allocation from byte-level segments to bit-level segments. Instead of allocating entire correcting bytes for each defective bit (conventional method), the invention segments the correcting bits into individual units that can be independently assigned to specific defective bits, reducing the total quantity of correcting bits needed.
Solution Approach 2:
The patent applies local quality by providing correction capability only where needed (at defective bit locations) rather than uniformly across entire bytes. Each correcting bit is locally assigned to replace a specific defective bit, creating a non-uniform distribution of correction resources that optimizes the ratio of correcting bits to protected data bits.
2Reliability
If entire bytes are replaced for a single defective bit (conventional method), then reliability is improved, but memory density decreases
Solution Approach 1:
The patent segments the replacement operation from byte-level to bit-level granularity. Instead of replacing entire bytes (8 bits) for a single defective bit, the invention replaces only the specific defective bit with a corresponding correcting bit, reducing the quantity of memory space consumed for correction from 8x to 1x the defective bit size.
Solution Approach 2:
The patent implements local quality by applying correction resources only to the specific location of defective bits rather than uniformly replacing entire bytes. This localized approach preserves the majority of functional bits within each byte, maintaining high memory density while ensuring reliability through targeted bit-level replacement.
3Reliability
If conventional correcting methods are used, then error correction capability is improved, but device complexity increases due to address management
Solution Approach 1:
The patent extracts the address management complexity from the correction system by using direct bit-position correspondence between defective bits and correcting bits. Instead of managing complex byte-level addresses and fuse array mappings, the invention extracts only the essential bit-position information needed for direct replacement, significantly simplifying the address management mechanism.
Solution Approach 2:
The patent inverts the conventional approach by having correcting bits take on the roles of data bits rather than having data bits replaced by complex correction codes. This inversion simplifies the system by using the same bit structure for both data and correction, eliminating the need for separate correction code management and address mapping tables.
Data Source
AI summary
A bit-scale memory correcting method comprises steps: providing a memory with a plurality of memory bytes each having M bits, wherein M is a positive integer; adding a correcting byte to each memory byte, wherein the correcting byte has N correcting bits, and wherein N is a positive integer and smaller than M; detecting whether there is any underperforming bit in all the memory bytes and correcting bytes; if no, terminating memory correction; if yes, using the non-underperforming bits of the correcting byte to replace the underperforming bits of the memory byte, wherein the quantity of the non-underperforming bits of the correcting byte is corresponding to the quantity of the underperforming bits of the memory byte. The present invention proposes a simple and fast memory bit correcting method to decrease the redundant bits for correcting memory bits.