Memory Page Read Window Reallocation for RBER-Limited Wordlines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Memory sub-systems face performance limitations due to the weakest wordline's narrow read window budgets (RWBs), leading to increased read errors and high Raw Bit Error Rate (RBER), which overwhelm error correction capabilities and compromise data integrity.
Innovation Solution
Implementing unbalanced RWBs by redistributing RWB from victim page types to benefactor page types and enhancing ECC parity bits for victim page types to manage higher RBER and maintain data integrity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If balanced RWBs are used across all page types, then read reliability is maintained, but memory performance is limited by the weakest wordline
Solution Approach 1:
The patent applies local quality by differentiating RWB allocation across different page types and wordlines. Instead of uniform RWB distribution, the system identifies specific page types (e.g., SLC vs. MLC pages) and wordlines with different performance characteristics, then allocates RWBs locally optimized for each category. This allows high-performance pages to use smaller RWBs while maintaining reliability for pages that need larger budgets.
Solution Approach 2:
The system dynamically changes the RWB parameter based on page type and wordline characteristics. By adjusting RWB sizes according to specific conditions (page type, wordline strength), the system optimizes the balance between read reliability and performance. Stronger wordlines use smaller RWBs for faster reads, while weaker wordlines use larger RWBs to maintain reliability.
2Reliability
If larger RWBs are used to improve read reliability, then read errors decrease, but read speed decreases due to wider voltage sweeps
Solution Approach 1:
Different page types receive locally optimized RWB sizes matched to their inherent reliability characteristics. SLC pages with higher natural reliability use smaller RWBs for faster reads, while MLC pages using smaller voltage margins receive larger RWBs to achieve the same error rates. This local optimization eliminates the need for uniform large RWBs across all pages.
Solution Approach 2:
The system applies partial RWB allocation - giving larger RWBs only to page types that need them (MLC, TLC) while using smaller RWBs for more reliable page types (SLC). This partial application of large RWBs achieves sufficient error correction for vulnerable pages without unnecessarily slowing down reads from more reliable pages.
3Productivity
If unbalanced RWBs are implemented, then memory performance improves, but some page types experience higher RBER
Solution Approach 1:
The system changes multiple parameters in coordination: RWB size, ECC code selection, and verify thresholds are all adjusted together based on page type. Victim page types receiving smaller RWBs are paired with stronger ECC protection and adjusted verify thresholds to compensate, maintaining acceptable RBER while enabling performance improvements on benefactor page types.
Solution Approach 2:
The solution combines multiple error protection mechanisms into a composite approach: unbalanced RWBs are combined with differential ECC allocation and adjusted verify thresholds. This composite strategy allows the system to tolerate higher RBER on victim pages through stronger ECC, while achieving overall performance gains from the unbalanced RWB distribution.
Data Source
AI summary
A processing device identifies a victim page type and a benefactor page type. The processing device selects a target wordline from a set of wordlines, wherein the target wordline is connected to a set of memory cells associated with the victim page type and the benefactor page type. The processing device allocates a portion of a read window budget (RWB) corresponding to the victim page type to the benefactor page type, wherein the RWB represents a margin between neighboring threshold voltage distributions of memory cells.