Memory Soft Data Determination via Log-Likelihood Ratios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Memory devices face challenges in efficiently determining soft data from hard reads due to the time-consuming nature of soft read operations, which impact memory throughput and require advanced error correction codes to handle higher bit error rates associated with multi-level cell configurations.
Innovation Solution
The method involves determining soft data from a hard read by analyzing the state of memory cells and using pre-determined probabilities and log-likelihood ratios (LLRs) to compute soft data, allowing for the determination of confidence levels without the need for additional data transfer, thereby improving memory device throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a soft read operation is performed to precisely determine voltage levels and generate soft data, then error correction capability is improved, but memory throughput deteriorates due to the time-consuming nature of soft read operations
Solution Approach 1:
The patent performs soft read operations in advance during programming phases or idle periods to generate and store soft data (LLRs) before actual read operations. This preliminary action allows the system to have soft data ready when needed, avoiding time-consuming soft reads during critical data access operations, thus maintaining both error correction capability and memory throughput
Solution Approach 2:
The patent extracts and stores only the essential soft data information (log-likelihood ratios) in a compact form during soft read operations, separating this data generation process from the main data access path. By extracting only the necessary confidence information and storing it efficiently, the system maintains error correction capability without the full overhead of repeated soft read operations
2Reliability
If soft read operations are performed to generate soft data with confidence levels, then error correction performance is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent transforms the continuous voltage measurement problem into a discrete log-likelihood ratio representation by changing the parameter domain. Instead of dealing with continuous voltage levels and complex probability calculations, the system converts measurements into LLR values that can be processed using simpler algebraic operations, reducing computational complexity while maintaining error correction performance
Solution Approach 2:
The patent replaces complex probabilistic calculations and continuous voltage analysis with algebraic operations on log-likelihood ratios. This substitution transforms the computational mechanism from complex statistical processing to simpler arithmetic operations, reducing the computational burden while preserving the essential error correction functionality
3Quantity of substance
If multi-level cell configurations are used to increase memory density, then storage capacity is improved, but bit error rates increase requiring more advanced error correction codes
Solution Approach 1:
The patent introduces log-likelihood ratios as an intermediary representation between the physical voltage states of multi-level cells and the logical data values. This intermediary provides confidence information about each bit decision, allowing error correction codes to operate more effectively on the ambiguous voltage levels inherent in MLC configurations, thereby maintaining reliability while enabling higher storage capacity
Data Source
AI summary
Apparatuses and methods involving the determination of soft data from hard reads are provided. One example method can include determining, using a hard read, a state of a memory cell. Soft data is determined based, at least partially, on the determined state.